diff --git "a/attnserver.run_attnserver.slurm.sh.343239.err.log" "b/attnserver.run_attnserver.slurm.sh.343239.err.log" --- "a/attnserver.run_attnserver.slurm.sh.343239.err.log" +++ "b/attnserver.run_attnserver.slurm.sh.343239.err.log" @@ -1672,3 +1672,5636 @@ W0621 22:08:33.247000 799671 site-packages/torch/distributed/run.py:766] W0621 22:08:33.247000 799671 site-packages/torch/distributed/run.py:766] ***************************************** W0621 22:08:33.247000 799671 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. W0621 22:08:33.247000 799671 site-packages/torch/distributed/run.py:766] ***************************************** +[rank2]:[W621 22:08:55.497229702 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:08:55.497239289 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank4]:[W621 22:08:55.497765256 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:08:55.498093266 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:08:55.498107467 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:08:55.498183197 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:08:55.498336978 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:08:55.920520655 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:08:55.926945425 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:08:55.927018126 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:08:55.927032304 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:08:55.927055811 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:08:55.927064585 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:08:55.927228751 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:08:55.016786338 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:08:55.631087831 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank1]:[W621 22:10:01.451704685 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank0]:[W621 22:10:01.544310900 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank8]:[W621 22:10:01.130709113 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:10:01.139504332 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:10:01.179912808 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:10:01.809400808 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:10:01.809491536 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:10:01.819587726 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank10]:[W621 22:10:01.241455197 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:10:02.532755463 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank4]:[W621 22:10:02.112174248 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank12]:[W621 22:10:02.533684183 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank6]:[W621 22:10:02.303905094 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:10:02.734858853 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank2]:[W621 22:10:02.354336719 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank14]:[W621 22:10:02.805980035 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=24576 ++ PROF_CTX_LENGTH=24576 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L24576*tp2.cp8.bs4.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L24576*tp2.cp8.bs4.json' ']' ++ echo 'Running ctx_length=24576, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=4' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 24576 --max-position-embeddings 24576 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 24576 --max-position-embeddings 24576 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:10:09.742000 2148954 site-packages/torch/distributed/run.py:766] +W0621 22:10:09.742000 2148954 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:10:09.742000 2148954 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:10:09.742000 2148954 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:10:09.847000 803182 site-packages/torch/distributed/run.py:766] +W0621 22:10:09.847000 803182 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:10:09.847000 803182 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:10:09.847000 803182 site-packages/torch/distributed/run.py:766] ***************************************** +[rank5]:[W621 22:10:32.734720374 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:10:32.734720556 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank4]:[W621 22:10:32.734779534 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:10:32.735546172 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:10:32.157485497 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:10:32.735610462 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:10:32.736159796 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:10:32.741413692 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:10:32.163754172 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:10:32.163762158 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:10:32.163844717 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:10:32.163897206 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:10:32.164516213 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:10:32.167801018 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:10:32.253730477 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:10:32.867427299 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank0]:[W621 22:12:23.767095493 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:12:23.242640599 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:12:23.263268669 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank1]:[W621 22:12:23.927843666 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:12:23.953081376 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:12:23.963024825 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:12:23.434782178 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank12]:[W621 22:12:24.466393554 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:12:24.063992109 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank10]:[W621 22:12:24.557009247 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank14]:[W621 22:12:24.577265220 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank2]:[W621 22:12:24.166596376 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank4]:[W621 22:12:24.176295123 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank6]:[W621 22:12:24.447926659 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:12:24.938488092 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank8]:[W621 22:12:24.040788010 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=32768 ++ PROF_CTX_LENGTH=32768 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L32768*tp2.cp8.bs4.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L32768*tp2.cp8.bs4.json' ']' ++ echo 'Running ctx_length=32768, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=4' ++ srun bash ./attnserver.sh ++ which python3 ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 32768 --max-position-embeddings 32768 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 32768 --max-position-embeddings 32768 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:12:34.689000 806714 site-packages/torch/distributed/run.py:766] +W0621 22:12:34.689000 806714 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:12:34.689000 806714 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:12:34.689000 806714 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:12:34.705000 2152539 site-packages/torch/distributed/run.py:766] +W0621 22:12:34.705000 2152539 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:12:34.705000 2152539 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:12:34.705000 2152539 site-packages/torch/distributed/run.py:766] ***************************************** +[rank2]:[W621 22:12:57.392208499 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank4]:[W621 22:12:57.393162117 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:12:57.393218736 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:12:57.821935688 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:12:57.822011850 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:12:57.822340997 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:12:57.403613800 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:12:57.403648697 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:12:57.403795477 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:12:57.404858992 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:12:57.832423908 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:12:57.832479269 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:12:57.832634553 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:12:57.833192566 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:12:57.913902306 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:12:57.531022454 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 58.95 GiB is free. Including non-PyTorch memory, this process has 80.85 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 907.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 58.82 GiB is free. Including non-PyTorch memory, this process has 80.96 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 1.01 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 59.34 GiB is free. Including non-PyTorch memory, this process has 80.43 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 1.01 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 59.33 GiB is free. Including non-PyTorch memory, this process has 80.47 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 523.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 59.84 GiB is free. Including non-PyTorch memory, this process has 79.93 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 523.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 60.12 GiB is free. Including non-PyTorch memory, this process has 79.68 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 267.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 59.83 GiB is free. Including non-PyTorch memory, this process has 79.98 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 31.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 59.59 GiB is free. Including non-PyTorch memory, this process has 80.21 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 267.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 60.08 GiB is free. Including non-PyTorch memory, this process has 79.70 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 267.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 59.79 GiB is free. Including non-PyTorch memory, this process has 79.99 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 31.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 60.12 GiB is free. Including non-PyTorch memory, this process has 79.68 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 267.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 59.83 GiB is free. Including non-PyTorch memory, this process has 79.95 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 523.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 59.56 GiB is free. Including non-PyTorch memory, this process has 80.22 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 267.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 59.85 GiB is free. Including non-PyTorch memory, this process has 79.95 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 523.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 59.32 GiB is free. Including non-PyTorch memory, this process has 80.46 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 523.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 59.85 GiB is free. Including non-PyTorch memory, this process has 79.95 GiB memory in use. Of the allocated memory 75.36 GiB is allocated by PyTorch, and 523.79 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]:[W621 22:13:44.901084087 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:13:45.132754782 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:13:45.956660455 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:13:45.555909675 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:13:45.037785984 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:13:45.047754088 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:13:45.808235126 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:13:45.300067945 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:13:47.335000 2152539 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2152610 closing signal SIGTERM +W0621 22:13:47.338000 2152539 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2152612 closing signal SIGTERM +W0621 22:13:47.343000 2152539 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2152613 closing signal SIGTERM +W0621 22:13:47.343000 2152539 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2152614 closing signal SIGTERM +W0621 22:13:47.346000 2152539 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2152616 closing signal SIGTERM +W0621 22:13:47.366000 2152539 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2152617 closing signal SIGTERM +W0621 22:13:48.020000 806714 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 806785 closing signal SIGTERM +W0621 22:13:48.024000 806714 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 806787 closing signal SIGTERM +W0621 22:13:48.026000 806714 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 806788 closing signal SIGTERM +W0621 22:13:48.027000 806714 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 806789 closing signal SIGTERM +W0621 22:13:48.028000 806714 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 806790 closing signal SIGTERM +W0621 22:13:48.029000 806714 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 806791 closing signal SIGTERM +W0621 22:13:48.032000 806714 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 806792 closing signal SIGTERM +E0621 22:13:52.661000 2152539 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 2152611) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +E0621 22:13:52.678000 806714 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 806786) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:13:47 + host : fs-mbz-gpu-141 + rank : 5 (local_rank: 5) + exitcode : 1 (pid: 2152615) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:13:47 + host : fs-mbz-gpu-141 + rank : 1 (local_rank: 1) + exitcode : 1 (pid: 2152611) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:13:48 + host : fs-mbz-gpu-188 + rank : 9 (local_rank: 1) + exitcode : 1 (pid: 806786) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=40960 ++ PROF_CTX_LENGTH=40960 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L40960*tp2.cp8.bs4.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L40960*tp2.cp8.bs4.json' ']' ++ echo 'Running ctx_length=40960, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=4' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 40960 --max-position-embeddings 40960 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 40960 --max-position-embeddings 40960 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:13:55.804000 809683 site-packages/torch/distributed/run.py:766] +W0621 22:13:55.804000 809683 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:13:55.804000 809683 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:13:55.804000 809683 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:13:55.806000 2155549 site-packages/torch/distributed/run.py:766] +W0621 22:13:55.806000 2155549 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:13:55.806000 2155549 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:13:55.806000 2155549 site-packages/torch/distributed/run.py:766] ***************************************** +[rank1]:[W621 22:14:19.537212229 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:14:19.537236205 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:14:19.537250857 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:14:19.537287136 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:14:19.960135281 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:14:19.960317672 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:14:19.960331362 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:14:19.960465197 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:14:19.543360966 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank4]:[W621 22:14:19.543461170 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:14:19.546531533 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:14:19.968662031 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:14:19.968709240 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:14:19.969473081 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:14:19.052200954 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:14:19.674898383 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank2]: output_tensor = model( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank2]: return self._call_impl(*args, **kwargs) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank2]: return forward_call(*args, **kwargs) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank2]: return self.module(*inputs, **kwargs) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank2]: return self._call_impl(*args, **kwargs) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank2]: return forward_call(*args, **kwargs) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank2]: outputs = self.module(*inputs, **kwargs) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank2]: return self._call_impl(*args, **kwargs) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank2]: return forward_call(*args, **kwargs) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank2]: loss = self.compute_language_model_loss(labels, logits) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank2]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank2]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank2]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank2]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank2]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 2 has a total capacity of 139.81 GiB of which 1.94 GiB is free. Including non-PyTorch memory, this process has 137.82 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 859.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank3]: output_tensor = model( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank3]: return self._call_impl(*args, **kwargs) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank3]: return forward_call(*args, **kwargs) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank3]: return self.module(*inputs, **kwargs) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank3]: return self._call_impl(*args, **kwargs) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank3]: return forward_call(*args, **kwargs) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank3]: outputs = self.module(*inputs, **kwargs) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank3]: return self._call_impl(*args, **kwargs) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank3]: return forward_call(*args, **kwargs) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank3]: loss = self.compute_language_model_loss(labels, logits) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank3]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank3]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank3]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank3]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank3]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 3 has a total capacity of 139.81 GiB of which 2.44 GiB is free. Including non-PyTorch memory, this process has 137.37 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 379.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank4]: output_tensor = model( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank4]: return self._call_impl(*args, **kwargs) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank4]: return forward_call(*args, **kwargs) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank4]: return self.module(*inputs, **kwargs) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank4]: return self._call_impl(*args, **kwargs) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank4]: return forward_call(*args, **kwargs) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank4]: outputs = self.module(*inputs, **kwargs) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank4]: return self._call_impl(*args, **kwargs) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank4]: return forward_call(*args, **kwargs) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank4]: loss = self.compute_language_model_loss(labels, logits) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank4]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank4]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank4]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank4]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank4]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 4 has a total capacity of 139.81 GiB of which 1.89 GiB is free. Including non-PyTorch memory, this process has 137.87 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 909.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank0]: output_tensor = model( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank0]: return self._call_impl(*args, **kwargs) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank0]: return forward_call(*args, **kwargs) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank0]: return self.module(*inputs, **kwargs) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank0]: return self._call_impl(*args, **kwargs) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank0]: return forward_call(*args, **kwargs) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank0]: outputs = self.module(*inputs, **kwargs) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank0]: return self._call_impl(*args, **kwargs) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank0]: return forward_call(*args, **kwargs) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank0]: loss = self.compute_language_model_loss(labels, logits) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank0]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank0]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank0]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank0]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank0]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 0 has a total capacity of 139.81 GiB of which 2.51 GiB is free. Including non-PyTorch memory, this process has 137.24 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 749.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank5]: output_tensor = model( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank5]: return self._call_impl(*args, **kwargs) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank5]: return forward_call(*args, **kwargs) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank5]: return self.module(*inputs, **kwargs) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank5]: return self._call_impl(*args, **kwargs) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank5]: return forward_call(*args, **kwargs) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank5]: outputs = self.module(*inputs, **kwargs) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank5]: return self._call_impl(*args, **kwargs) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank5]: return forward_call(*args, **kwargs) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank5]: loss = self.compute_language_model_loss(labels, logits) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank5]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank5]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank5]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank5]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank5]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 5 has a total capacity of 139.81 GiB of which 2.08 GiB is free. Including non-PyTorch memory, this process has 137.73 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 749.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank1]: output_tensor = model( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank1]: return self._call_impl(*args, **kwargs) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank1]: return forward_call(*args, **kwargs) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank1]: return self.module(*inputs, **kwargs) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank1]: return self._call_impl(*args, **kwargs) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank1]: return forward_call(*args, **kwargs) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank1]: outputs = self.module(*inputs, **kwargs) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank1]: return self._call_impl(*args, **kwargs) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank1]: return forward_call(*args, **kwargs) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank1]: loss = self.compute_language_model_loss(labels, logits) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank1]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank1]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank1]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank1]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank1]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 1 has a total capacity of 139.81 GiB of which 2.86 GiB is free. Including non-PyTorch memory, this process has 136.95 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 429.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank7]: output_tensor = model( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank7]: return self._call_impl(*args, **kwargs) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank7]: return forward_call(*args, **kwargs) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank7]: return self.module(*inputs, **kwargs) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank7]: return self._call_impl(*args, **kwargs) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank7]: return forward_call(*args, **kwargs) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank7]: outputs = self.module(*inputs, **kwargs) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank7]: return self._call_impl(*args, **kwargs) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank7]: return forward_call(*args, **kwargs) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank7]: loss = self.compute_language_model_loss(labels, logits) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank7]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank7]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank7]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank7]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank7]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 7 has a total capacity of 139.81 GiB of which 2.29 GiB is free. Including non-PyTorch memory, this process has 137.51 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 749.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank6]: output_tensor = model( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank6]: return self._call_impl(*args, **kwargs) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank6]: return forward_call(*args, **kwargs) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank6]: return self.module(*inputs, **kwargs) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank6]: return self._call_impl(*args, **kwargs) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank6]: return forward_call(*args, **kwargs) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank6]: outputs = self.module(*inputs, **kwargs) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank6]: return self._call_impl(*args, **kwargs) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank6]: return forward_call(*args, **kwargs) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank6]: loss = self.compute_language_model_loss(labels, logits) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank6]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank6]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank6]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank6]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank6]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 6 has a total capacity of 139.81 GiB of which 2.26 GiB is free. Including non-PyTorch memory, this process has 137.49 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 749.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank14]: output_tensor = model( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank14]: return self._call_impl(*args, **kwargs) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank14]: return forward_call(*args, **kwargs) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank14]: return self.module(*inputs, **kwargs) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank14]: return self._call_impl(*args, **kwargs) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank14]: return forward_call(*args, **kwargs) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank14]: outputs = self.module(*inputs, **kwargs) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank14]: return self._call_impl(*args, **kwargs) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank14]: return forward_call(*args, **kwargs) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank14]: loss = self.compute_language_model_loss(labels, logits) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank14]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank14]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank14]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank14]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank14]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 6 has a total capacity of 139.81 GiB of which 1.78 GiB is free. Including non-PyTorch memory, this process has 137.98 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 1.20 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank15]: output_tensor = model( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank15]: return self._call_impl(*args, **kwargs) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank15]: return forward_call(*args, **kwargs) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank15]: return self.module(*inputs, **kwargs) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank15]: return self._call_impl(*args, **kwargs) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank15]: return forward_call(*args, **kwargs) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank15]: outputs = self.module(*inputs, **kwargs) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank15]: return self._call_impl(*args, **kwargs) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank15]: return forward_call(*args, **kwargs) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank15]: loss = self.compute_language_model_loss(labels, logits) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank15]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank15]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank15]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank15]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank15]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 7 has a total capacity of 139.81 GiB of which 2.00 GiB is free. Including non-PyTorch memory, this process has 137.80 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 1.04 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank12]: output_tensor = model( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank12]: return self._call_impl(*args, **kwargs) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank12]: return forward_call(*args, **kwargs) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank12]: return self.module(*inputs, **kwargs) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank12]: return self._call_impl(*args, **kwargs) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank12]: return forward_call(*args, **kwargs) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank12]: outputs = self.module(*inputs, **kwargs) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank12]: return self._call_impl(*args, **kwargs) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank12]: return forward_call(*args, **kwargs) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank12]: loss = self.compute_language_model_loss(labels, logits) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank12]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank12]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank12]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank12]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank12]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 4 has a total capacity of 139.81 GiB of which 1.40 GiB is free. Including non-PyTorch memory, this process has 138.35 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 1.36 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank9]: output_tensor = model( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank9]: return self._call_impl(*args, **kwargs) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank9]: return forward_call(*args, **kwargs) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank9]: return self.module(*inputs, **kwargs) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank9]: return self._call_impl(*args, **kwargs) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank9]: return forward_call(*args, **kwargs) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank9]: outputs = self.module(*inputs, **kwargs) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank9]: return self._call_impl(*args, **kwargs) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank9]: return forward_call(*args, **kwargs) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank9]: loss = self.compute_language_model_loss(labels, logits) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank9]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank9]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank9]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank9]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank9]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 1 has a total capacity of 139.81 GiB of which 2.25 GiB is free. Including non-PyTorch memory, this process has 137.55 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 1.04 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank13]: output_tensor = model( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank13]: return self._call_impl(*args, **kwargs) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank13]: return forward_call(*args, **kwargs) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank13]: return self.module(*inputs, **kwargs) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank13]: return self._call_impl(*args, **kwargs) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank13]: return forward_call(*args, **kwargs) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank13]: outputs = self.module(*inputs, **kwargs) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank13]: return self._call_impl(*args, **kwargs) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank13]: return forward_call(*args, **kwargs) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank13]: loss = self.compute_language_model_loss(labels, logits) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank13]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank13]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank13]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank13]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank13]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 5 has a total capacity of 139.81 GiB of which 1.78 GiB is free. Including non-PyTorch memory, this process has 138.02 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 1.04 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank8]: output_tensor = model( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank8]: return self._call_impl(*args, **kwargs) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank8]: return forward_call(*args, **kwargs) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank8]: return self.module(*inputs, **kwargs) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank8]: return self._call_impl(*args, **kwargs) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank8]: return forward_call(*args, **kwargs) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank8]: outputs = self.module(*inputs, **kwargs) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank8]: return self._call_impl(*args, **kwargs) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank8]: return forward_call(*args, **kwargs) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank8]: loss = self.compute_language_model_loss(labels, logits) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank8]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank8]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank8]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank8]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank8]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 0 has a total capacity of 139.81 GiB of which 2.18 GiB is free. Including non-PyTorch memory, this process has 137.57 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 1.04 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank11]: output_tensor = model( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank11]: return self._call_impl(*args, **kwargs) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank11]: return forward_call(*args, **kwargs) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank11]: return self.module(*inputs, **kwargs) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank11]: return self._call_impl(*args, **kwargs) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank11]: return forward_call(*args, **kwargs) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank11]: outputs = self.module(*inputs, **kwargs) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank11]: return self._call_impl(*args, **kwargs) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank11]: return forward_call(*args, **kwargs) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank11]: loss = self.compute_language_model_loss(labels, logits) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank11]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank11]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank11]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank11]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank11]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 3 has a total capacity of 139.81 GiB of which 1.78 GiB is free. Including non-PyTorch memory, this process has 138.02 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 1.04 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 460, in forward_step +[rank10]: output_tensor = model( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank10]: return self._call_impl(*args, **kwargs) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank10]: return forward_call(*args, **kwargs) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/distributed/data_parallel_base.py", line 22, in forward +[rank10]: return self.module(*inputs, **kwargs) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank10]: return self._call_impl(*args, **kwargs) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank10]: return forward_call(*args, **kwargs) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/transformer/module.py", line 236, in forward +[rank10]: outputs = self.module(*inputs, **kwargs) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl +[rank10]: return self._call_impl(*args, **kwargs) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl +[rank10]: return forward_call(*args, **kwargs) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_model.py", line 414, in forward +[rank10]: loss = self.compute_language_model_loss(labels, logits) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/common/language_module/language_module.py", line 98, in compute_language_model_loss +[rank10]: loss = tensor_parallel.vocab_parallel_cross_entropy(logits, labels) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 232, in vocab_parallel_cross_entropy +[rank10]: return _VocabParallelCrossEntropy.apply(vocab_parallel_logits, target, label_smoothing) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/autograd/function.py", line 575, in apply +[rank10]: return super().apply(*args, **kwargs) # type: ignore[misc] +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 127, in forward +[rank10]: vocab_parallel_logits, logits_max = VocabParallelCrossEntropy.calculate_logits_max( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/tensor_parallel/cross_entropy.py", line 28, in calculate_logits_max +[rank10]: vocab_parallel_logits = vocab_parallel_logits.float() +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.70 GiB. GPU 2 has a total capacity of 139.81 GiB of which 1.56 GiB is free. Including non-PyTorch memory, this process has 138.20 GiB memory in use. Of the allocated memory 134.16 GiB is allocated by PyTorch, and 1.20 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +W0621 22:15:12.527000 809683 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 809755 closing signal SIGTERM +W0621 22:15:12.530000 809683 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 809756 closing signal SIGTERM +W0621 22:15:12.531000 809683 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 809757 closing signal SIGTERM +W0621 22:15:12.533000 809683 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 809758 closing signal SIGTERM +W0621 22:15:12.534000 809683 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 809759 closing signal SIGTERM +W0621 22:15:12.537000 809683 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 809760 closing signal SIGTERM +W0621 22:15:12.537000 809683 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 809761 closing signal SIGTERM +W0621 22:15:12.550000 2155549 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2155619 closing signal SIGTERM +W0621 22:15:12.555000 2155549 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2155620 closing signal SIGTERM +W0621 22:15:12.555000 2155549 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2155621 closing signal SIGTERM +W0621 22:15:12.557000 2155549 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2155623 closing signal SIGTERM +W0621 22:15:12.560000 2155549 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2155625 closing signal SIGTERM +W0621 22:15:12.562000 2155549 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2155626 closing signal SIGTERM +E0621 22:15:14.067000 809683 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 7 (pid: 809762) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:15:12 + host : fs-mbz-gpu-188 + rank : 15 (local_rank: 7) + exitcode : 1 (pid: 809762) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x +W0621 22:15:16.153000 2155549 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1341] The node 'fs-mbz-gpu-141_2155549_0' has failed to send a keep-alive heartbeat to the rendezvous '343239' due to an error of type RendezvousConnectionError. +E0621 22:15:18.762000 2155549 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 3 (pid: 2155622) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +[W621 22:15:18.775451284 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-141]:40626, remote=[fs-mbz-gpu-188]:29500): Broken pipe +Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first): +frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string, std::allocator >) + 0x98 (0x150d4bf785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so) +frame #1: + 0x5ba8afe (0x150d3525aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #2: + 0x5baa358 (0x150d3525c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #3: + 0x5babb3e (0x150d3525db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #4: c10d::TCPStore::doWait(c10::ArrayRef, std::allocator > >, std::chrono::duration >) + 0x1a6 (0x150d35257ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string, std::allocator > const&) + 0x33 (0x150d35257ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #6: c10d::TCPStore::get(std::__cxx11::basic_string, std::allocator > const&) + 0xab (0x150d35258f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #7: + 0xc0f526 (0x150d4458b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) +frame #8: + 0x37f17d (0x150d43cfb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) + +frame #26: + 0x29d90 (0x150d4d2d7d90 in /lib/x86_64-linux-gnu/libc.so.6) +frame #27: __libc_start_main + 0x80 (0x150d4d2d7e40 in /lib/x86_64-linux-gnu/libc.so.6) + +W0621 22:15:18.779000 2155549 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2155549_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +[W621 22:15:18.793506856 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-141]:40626, remote=[fs-mbz-gpu-188]:29500): Broken pipe +Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first): +frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string, std::allocator >) + 0x98 (0x150d4bf785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so) +frame #1: + 0x5ba8afe (0x150d3525aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #2: + 0x5baa358 (0x150d3525c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #3: + 0x5babb3e (0x150d3525db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #4: c10d::TCPStore::doWait(c10::ArrayRef, std::allocator > >, std::chrono::duration >) + 0x1a6 (0x150d35257ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string, std::allocator > const&) + 0x33 (0x150d35257ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #6: c10d::TCPStore::get(std::__cxx11::basic_string, std::allocator > const&) + 0xab (0x150d35258f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #7: + 0xc0f526 (0x150d4458b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) +frame #8: + 0x37f17d (0x150d43cfb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) + +frame #26: + 0x29d90 (0x150d4d2d7d90 in /lib/x86_64-linux-gnu/libc.so.6) +frame #27: __libc_start_main + 0x80 (0x150d4d2d7e40 in /lib/x86_64-linux-gnu/libc.so.6) + +W0621 22:15:18.791000 2155549 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2155549_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +[W621 22:15:18.803816022 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-141]:40626, remote=[fs-mbz-gpu-188]:29500): Broken pipe +Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first): +frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string, std::allocator >) + 0x98 (0x150d4bf785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so) +frame #1: + 0x5ba8afe (0x150d3525aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #2: + 0x5baa358 (0x150d3525c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #3: + 0x5babb3e (0x150d3525db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #4: c10d::TCPStore::doWait(c10::ArrayRef, std::allocator > >, std::chrono::duration >) + 0x1a6 (0x150d35257ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string, std::allocator > const&) + 0x33 (0x150d35257ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #6: c10d::TCPStore::get(std::__cxx11::basic_string, std::allocator > const&) + 0xab (0x150d35258f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #7: + 0xc0f526 (0x150d4458b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) +frame #8: + 0x37f17d (0x150d43cfb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) + +frame #26: + 0x29d90 (0x150d4d2d7d90 in /lib/x86_64-linux-gnu/libc.so.6) +frame #27: __libc_start_main + 0x80 (0x150d4d2d7e40 in /lib/x86_64-linux-gnu/libc.so.6) + +W0621 22:15:18.819000 2155549 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2155549_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:15:12 + host : fs-mbz-gpu-141 + rank : 5 (local_rank: 5) + exitcode : 1 (pid: 2155624) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:15:12 + host : fs-mbz-gpu-141 + rank : 3 (local_rank: 3) + exitcode : 1 (pid: 2155622) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=49152 ++ PROF_CTX_LENGTH=49152 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L49152*tp2.cp8.bs4.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L49152*tp2.cp8.bs4.json' ']' ++ echo 'Running ctx_length=49152, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=4' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 49152 --max-position-embeddings 49152 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 49152 --max-position-embeddings 49152 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:15:24.861000 812455 site-packages/torch/distributed/run.py:766] +W0621 22:15:24.861000 812455 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:15:24.861000 812455 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:15:24.861000 812455 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:15:24.985000 2158376 site-packages/torch/distributed/run.py:766] +W0621 22:15:24.985000 2158376 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:15:24.985000 2158376 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:15:24.985000 2158376 site-packages/torch/distributed/run.py:766] ***************************************** +[rank4]:[W621 22:15:48.217183036 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:15:48.217694511 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:15:48.217711601 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:15:48.224793696 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:15:48.224797114 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:15:48.225418797 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:15:48.226557646 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:15:48.651017548 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:15:48.651091898 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:15:48.651097762 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:15:48.651157014 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:15:48.651189658 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:15:48.651422823 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:15:48.651919476 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:15:48.749370196 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:15:48.351495883 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 134.53 GiB is free. Including non-PyTorch memory, this process has 5.27 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 134.53 GiB is free. Including non-PyTorch memory, this process has 5.27 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 134.51 GiB is free. Including non-PyTorch memory, this process has 5.29 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 134.51 GiB is free. Including non-PyTorch memory, this process has 5.29 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 134.51 GiB is free. Including non-PyTorch memory, this process has 5.29 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 134.53 GiB is free. Including non-PyTorch memory, this process has 5.27 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 134.53 GiB is free. Including non-PyTorch memory, this process has 5.27 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 134.53 GiB is free. Including non-PyTorch memory, this process has 5.27 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 134.51 GiB is free. Including non-PyTorch memory, this process has 5.29 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 134.51 GiB is free. Including non-PyTorch memory, this process has 5.29 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 134.51 GiB is free. Including non-PyTorch memory, this process has 5.29 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 134.51 GiB is free. Including non-PyTorch memory, this process has 5.29 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 134.53 GiB is free. Including non-PyTorch memory, this process has 5.27 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 134.51 GiB is free. Including non-PyTorch memory, this process has 5.29 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 134.53 GiB is free. Including non-PyTorch memory, this process has 5.27 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 144.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 134.53 GiB is free. Including non-PyTorch memory, this process has 5.27 GiB memory in use. Of the allocated memory 3.60 GiB is allocated by PyTorch, and 200.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]:[W621 22:15:59.383004615 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:15:59.839683676 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:15:59.438959696 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:15:59.890333822 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:15:59.910473467 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:15:59.930669121 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:15:59.580048715 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:15:59.590230720 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:16:01.038000 812455 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 812527 closing signal SIGTERM +W0621 22:16:01.040000 812455 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 812529 closing signal SIGTERM +W0621 22:16:01.041000 2158376 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2158446 closing signal SIGTERM +W0621 22:16:01.043000 812455 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 812531 closing signal SIGTERM +W0621 22:16:01.046000 812455 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 812532 closing signal SIGTERM +W0621 22:16:01.047000 812455 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 812533 closing signal SIGTERM +W0621 22:16:01.045000 2158376 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2158448 closing signal SIGTERM +W0621 22:16:01.047000 2158376 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2158450 closing signal SIGTERM +W0621 22:16:01.049000 2158376 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2158451 closing signal SIGTERM +W0621 22:16:01.050000 2158376 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2158452 closing signal SIGTERM +W0621 22:16:01.052000 2158376 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2158453 closing signal SIGTERM +E0621 22:16:01.190000 812455 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 812528) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:16:01 + host : fs-mbz-gpu-188 + rank : 11 (local_rank: 3) + exitcode : 1 (pid: 812530) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +[2]: + time : 2025-06-21_22:16:01 + host : fs-mbz-gpu-188 + rank : 15 (local_rank: 7) + exitcode : 1 (pid: 812534) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:16:01 + host : fs-mbz-gpu-188 + rank : 9 (local_rank: 1) + exitcode : 1 (pid: 812528) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ +E0621 22:16:01.492000 2158376 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 2158447) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +W0621 22:16:01.508000 2158376 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2158376_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +[W621 22:16:01.521789122 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-141]:36432, remote=[fs-mbz-gpu-188]:29500): Broken pipe +Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first): +frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string, std::allocator >) + 0x98 (0x15408f3785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so) +frame #1: + 0x5ba8afe (0x15407825aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #2: + 0x5baa358 (0x15407825c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #3: + 0x5babb3e (0x15407825db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #4: c10d::TCPStore::doWait(c10::ArrayRef, std::allocator > >, std::chrono::duration >) + 0x1a6 (0x154078257ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string, std::allocator > const&) + 0x33 (0x154078257ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #6: c10d::TCPStore::get(std::__cxx11::basic_string, std::allocator > const&) + 0xab (0x154078258f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #7: + 0xc0f526 (0x15408758b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) +frame #8: + 0x37f17d (0x154086cfb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) + +frame #26: + 0x29d90 (0x1540903dfd90 in /lib/x86_64-linux-gnu/libc.so.6) +frame #27: __libc_start_main + 0x80 (0x1540903dfe40 in /lib/x86_64-linux-gnu/libc.so.6) + +W0621 22:16:01.522000 2158376 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2158376_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +[W621 22:16:01.535233232 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-141]:36432, remote=[fs-mbz-gpu-188]:29500): Broken pipe +Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first): +frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string, std::allocator >) + 0x98 (0x15408f3785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so) +frame #1: + 0x5ba8afe (0x15407825aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #2: + 0x5baa358 (0x15407825c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #3: + 0x5babb3e (0x15407825db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #4: c10d::TCPStore::doWait(c10::ArrayRef, std::allocator > >, std::chrono::duration >) + 0x1a6 (0x154078257ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string, std::allocator > const&) + 0x33 (0x154078257ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #6: c10d::TCPStore::get(std::__cxx11::basic_string, std::allocator > const&) + 0xab (0x154078258f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #7: + 0xc0f526 (0x15408758b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) +frame #8: + 0x37f17d (0x154086cfb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) + +frame #26: + 0x29d90 (0x1540903dfd90 in /lib/x86_64-linux-gnu/libc.so.6) +frame #27: __libc_start_main + 0x80 (0x1540903dfe40 in /lib/x86_64-linux-gnu/libc.so.6) + +W0621 22:16:01.533000 2158376 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2158376_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:16:01 + host : fs-mbz-gpu-141 + rank : 3 (local_rank: 3) + exitcode : 1 (pid: 2158449) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:16:01 + host : fs-mbz-gpu-141 + rank : 1 (local_rank: 1) + exitcode : 1 (pid: 2158447) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=65536 ++ PROF_CTX_LENGTH=65536 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L65536*tp2.cp8.bs4.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L65536*tp2.cp8.bs4.json' ']' ++ echo 'Running ctx_length=65536, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=4' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 65536 --max-position-embeddings 65536 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 65536 --max-position-embeddings 65536 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:16:05.786000 814283 site-packages/torch/distributed/run.py:766] +W0621 22:16:05.786000 814283 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:16:05.786000 814283 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:16:05.786000 814283 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:16:05.800000 2160217 site-packages/torch/distributed/run.py:766] +W0621 22:16:05.800000 2160217 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:16:05.800000 2160217 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:16:05.800000 2160217 site-packages/torch/distributed/run.py:766] ***************************************** +[rank4]:[W621 22:16:28.565010314 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:16:28.565015971 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:16:28.565739073 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:16:28.574394971 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:16:28.575097583 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:16:28.997717565 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:16:28.576875382 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:16:28.998803704 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:16:28.998814585 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:16:28.998852645 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:16:28.998870690 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:16:28.998900750 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:16:28.999332090 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:16:28.578178468 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:16:28.100298034 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:16:28.695845166 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 133.76 GiB is free. Including non-PyTorch memory, this process has 6.04 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 133.76 GiB is free. Including non-PyTorch memory, this process has 6.04 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 133.76 GiB is free. Including non-PyTorch memory, this process has 6.04 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 133.78 GiB is free. Including non-PyTorch memory, this process has 6.02 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 133.76 GiB is free. Including non-PyTorch memory, this process has 6.04 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 133.78 GiB is free. Including non-PyTorch memory, this process has 6.02 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 133.76 GiB is free. Including non-PyTorch memory, this process has 6.04 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 133.76 GiB is free. Including non-PyTorch memory, this process has 6.04 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 133.78 GiB is free. Including non-PyTorch memory, this process has 6.02 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 133.78 GiB is free. Including non-PyTorch memory, this process has 6.02 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 133.78 GiB is free. Including non-PyTorch memory, this process has 6.02 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 133.78 GiB is free. Including non-PyTorch memory, this process has 6.02 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 133.78 GiB is free. Including non-PyTorch memory, this process has 6.02 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 133.78 GiB is free. Including non-PyTorch memory, this process has 6.02 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 133.76 GiB is free. Including non-PyTorch memory, this process has 6.04 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 133.76 GiB is free. Including non-PyTorch memory, this process has 6.04 GiB memory in use. Of the allocated memory 4.10 GiB is allocated by PyTorch, and 451.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]:[W621 22:16:40.677477945 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:16:40.761591557 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:16:40.233744344 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:16:40.294140840 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:16:40.294202284 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:16:40.354642274 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:16:40.963760159 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:16:40.004087718 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:16:42.236000 2160217 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2160288 closing signal SIGTERM +W0621 22:16:42.239000 2160217 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2160289 closing signal SIGTERM +W0621 22:16:42.240000 2160217 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2160290 closing signal SIGTERM +W0621 22:16:42.240000 2160217 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2160292 closing signal SIGTERM +W0621 22:16:42.245000 2160217 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2160293 closing signal SIGTERM +W0621 22:16:42.246000 2160217 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2160294 closing signal SIGTERM +W0621 22:16:42.248000 2160217 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2160295 closing signal SIGTERM +W0621 22:16:42.268000 814283 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 814353 closing signal SIGTERM +W0621 22:16:42.272000 814283 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 814355 closing signal SIGTERM +W0621 22:16:42.274000 814283 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 814356 closing signal SIGTERM +W0621 22:16:42.275000 814283 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 814357 closing signal SIGTERM +W0621 22:16:42.277000 814283 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 814358 closing signal SIGTERM +W0621 22:16:42.277000 814283 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 814359 closing signal SIGTERM +W0621 22:16:42.290000 814283 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 814360 closing signal SIGTERM +E0621 22:16:42.722000 814283 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 814354) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:16:42 + host : fs-mbz-gpu-188 + rank : 9 (local_rank: 1) + exitcode : 1 (pid: 814354) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ +E0621 22:16:42.840000 2160217 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 3 (pid: 2160291) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +[W621 22:16:42.867269209 TCPStore.cpp:115] [c10d] recvVector failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-141]:51834, remote=[fs-mbz-gpu-188]:29500): failed to recv, got 0 bytes +Exception raised from recvBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:678 (most recent call first): +frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string, std::allocator >) + 0x98 (0x149a493785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so) +frame #1: + 0x5ba8afe (0x149a3225aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #2: + 0x5baa0d0 (0x149a3225c0d0 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #3: + 0x5baa81d (0x149a3225c81d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #4: + 0x5bab4a9 (0x149a3225d4a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #5: c10d::TCPStore::compareSet(std::__cxx11::basic_string, std::allocator > const&, std::vector > const&, std::vector > const&) + 0x1fb (0x149a322574cb in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #6: + 0xc0f919 (0x149a4158b919 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) +frame #7: + 0x37f17d (0x149a40cfb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) + +frame #25: + 0x29d90 (0x149a4a3dad90 in /lib/x86_64-linux-gnu/libc.so.6) +frame #26: __libc_start_main + 0x80 (0x149a4a3dae40 in /lib/x86_64-linux-gnu/libc.so.6) + +W0621 22:16:42.866000 2160217 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2160217_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +[W621 22:16:42.880071780 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-141]:51834, remote=[fs-mbz-gpu-188]:29500): Broken pipe +Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first): +frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string, std::allocator >) + 0x98 (0x149a493785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so) +frame #1: + 0x5ba8afe (0x149a3225aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #2: + 0x5baa358 (0x149a3225c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #3: + 0x5babb3e (0x149a3225db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #4: c10d::TCPStore::compareSet(std::__cxx11::basic_string, std::allocator > const&, std::vector > const&, std::vector > const&) + 0x299 (0x149a32257569 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #5: + 0xc0f919 (0x149a4158b919 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) +frame #6: + 0x37f17d (0x149a40cfb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) + +frame #24: + 0x29d90 (0x149a4a3dad90 in /lib/x86_64-linux-gnu/libc.so.6) +frame #25: __libc_start_main + 0x80 (0x149a4a3dae40 in /lib/x86_64-linux-gnu/libc.so.6) + +W0621 22:16:42.877000 2160217 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2160217_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +[W621 22:16:42.889743565 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-141]:51834, remote=[fs-mbz-gpu-188]:29500): Broken pipe +Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first): +frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string, std::allocator >) + 0x98 (0x149a493785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so) +frame #1: + 0x5ba8afe (0x149a3225aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #2: + 0x5baa358 (0x149a3225c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #3: + 0x5babb3e (0x149a3225db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #4: c10d::TCPStore::compareSet(std::__cxx11::basic_string, std::allocator > const&, std::vector > const&, std::vector > const&) + 0x299 (0x149a32257569 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #5: + 0xc0f919 (0x149a4158b919 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) +frame #6: + 0x37f17d (0x149a40cfb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) + +frame #24: + 0x29d90 (0x149a4a3dad90 in /lib/x86_64-linux-gnu/libc.so.6) +frame #25: __libc_start_main + 0x80 (0x149a4a3dae40 in /lib/x86_64-linux-gnu/libc.so.6) + +W0621 22:16:42.886000 2160217 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2160217_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:16:42 + host : fs-mbz-gpu-141 + rank : 3 (local_rank: 3) + exitcode : 1 (pid: 2160291) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=81920 ++ PROF_CTX_LENGTH=81920 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L81920*tp2.cp8.bs4.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L81920*tp2.cp8.bs4.json' ']' ++ echo 'Running ctx_length=81920, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=4' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 81920 --max-position-embeddings 81920 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 81920 --max-position-embeddings 81920 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:16:47.111000 816091 site-packages/torch/distributed/run.py:766] +W0621 22:16:47.111000 816091 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:16:47.111000 816091 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:16:47.111000 816091 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:16:47.113000 2162076 site-packages/torch/distributed/run.py:766] +W0621 22:16:47.113000 2162076 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:16:47.113000 2162076 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:16:47.113000 2162076 site-packages/torch/distributed/run.py:766] ***************************************** +[rank4]:[W621 22:17:09.564362910 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:17:09.564369807 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:17:09.564437301 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:17:09.990064116 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:17:09.990661247 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:17:09.990880720 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:17:09.991719303 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:17:09.571717306 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:17:09.572255719 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:17:09.572885914 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:17:09.574957599 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:17:09.002429480 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:17:09.002489615 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:17:09.002673010 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:17:09.094832014 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:17:09.696840129 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 133.46 GiB is free. Including non-PyTorch memory, this process has 6.35 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 133.46 GiB is free. Including non-PyTorch memory, this process has 6.35 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 133.46 GiB is free. Including non-PyTorch memory, this process has 6.35 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 133.47 GiB is free. Including non-PyTorch memory, this process has 6.33 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 133.46 GiB is free. Including non-PyTorch memory, this process has 6.35 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 133.47 GiB is free. Including non-PyTorch memory, this process has 6.33 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 133.47 GiB is free. Including non-PyTorch memory, this process has 6.33 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 133.47 GiB is free. Including non-PyTorch memory, this process has 6.33 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 133.46 GiB is free. Including non-PyTorch memory, this process has 6.35 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 133.46 GiB is free. Including non-PyTorch memory, this process has 6.35 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 133.47 GiB is free. Including non-PyTorch memory, this process has 6.33 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 133.47 GiB is free. Including non-PyTorch memory, this process has 6.33 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 133.47 GiB is free. Including non-PyTorch memory, this process has 6.33 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 133.47 GiB is free. Including non-PyTorch memory, this process has 6.33 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 133.46 GiB is free. Including non-PyTorch memory, this process has 6.35 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 400.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 133.46 GiB is free. Including non-PyTorch memory, this process has 6.35 GiB memory in use. Of the allocated memory 4.61 GiB is allocated by PyTorch, and 248.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]:[W621 22:17:22.388842190 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:17:22.388863640 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:17:22.389142739 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:17:22.830612190 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:17:22.860870738 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:17:22.871020963 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:17:22.871086753 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank1]:[W621 22:17:22.616161510 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:17:23.954000 816091 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 816162 closing signal SIGTERM +W0621 22:17:23.961000 816091 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 816163 closing signal SIGTERM +W0621 22:17:23.962000 816091 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 816164 closing signal SIGTERM +W0621 22:17:23.964000 816091 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 816165 closing signal SIGTERM +W0621 22:17:23.964000 816091 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 816166 closing signal SIGTERM +W0621 22:17:23.966000 816091 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 816167 closing signal SIGTERM +W0621 22:17:23.966000 816091 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 816168 closing signal SIGTERM +W0621 22:17:24.002000 2162076 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2162147 closing signal SIGTERM +W0621 22:17:24.004000 2162076 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2162148 closing signal SIGTERM +W0621 22:17:24.004000 2162076 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2162149 closing signal SIGTERM +W0621 22:17:24.006000 2162076 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2162151 closing signal SIGTERM +W0621 22:17:24.010000 2162076 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2162153 closing signal SIGTERM +E0621 22:17:24.254000 2162076 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 3 (pid: 2162150) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:17:24 + host : fs-mbz-gpu-141 + rank : 5 (local_rank: 5) + exitcode : 1 (pid: 2162152) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +[2]: + time : 2025-06-21_22:17:24 + host : fs-mbz-gpu-141 + rank : 7 (local_rank: 7) + exitcode : 1 (pid: 2162154) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:17:24 + host : fs-mbz-gpu-141 + rank : 3 (local_rank: 3) + exitcode : 1 (pid: 2162150) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ +E0621 22:17:24.460000 816091 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 7 (pid: 816169) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:17:23 + host : fs-mbz-gpu-188 + rank : 15 (local_rank: 7) + exitcode : 1 (pid: 816169) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=98304 ++ PROF_CTX_LENGTH=98304 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L98304*tp2.cp8.bs4.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L98304*tp2.cp8.bs4.json' ']' ++ echo 'Running ctx_length=98304, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=4' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 98304 --max-position-embeddings 98304 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 98304 --max-position-embeddings 98304 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:17:28.849000 817940 site-packages/torch/distributed/run.py:766] +W0621 22:17:28.849000 817940 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:17:28.849000 817940 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:17:28.849000 817940 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:17:28.880000 2163941 site-packages/torch/distributed/run.py:766] +W0621 22:17:28.880000 2163941 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:17:28.880000 2163941 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:17:28.880000 2163941 site-packages/torch/distributed/run.py:766] ***************************************** +[rank4]:[W621 22:17:51.077966573 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:17:51.078949123 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:17:51.079259750 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:17:51.507233833 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:17:51.507273263 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:17:51.507370374 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:17:51.507429415 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:17:51.507921284 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:17:51.087232591 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:17:51.508439274 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:17:51.087255940 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:17:51.087264675 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:17:51.087402764 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:17:51.521251888 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:17:51.597862092 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:17:51.238481932 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 132.90 GiB is free. Including non-PyTorch memory, this process has 6.90 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 132.90 GiB is free. Including non-PyTorch memory, this process has 6.90 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 132.90 GiB is free. Including non-PyTorch memory, this process has 6.90 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 132.92 GiB is free. Including non-PyTorch memory, this process has 6.89 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 132.92 GiB is free. Including non-PyTorch memory, this process has 6.89 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 132.90 GiB is free. Including non-PyTorch memory, this process has 6.90 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 132.92 GiB is free. Including non-PyTorch memory, this process has 6.89 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 132.92 GiB is free. Including non-PyTorch memory, this process has 6.89 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 132.92 GiB is free. Including non-PyTorch memory, this process has 6.89 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 132.90 GiB is free. Including non-PyTorch memory, this process has 6.90 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 132.92 GiB is free. Including non-PyTorch memory, this process has 6.89 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 132.90 GiB is free. Including non-PyTorch memory, this process has 6.90 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 132.92 GiB is free. Including non-PyTorch memory, this process has 6.89 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 132.90 GiB is free. Including non-PyTorch memory, this process has 6.90 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 132.90 GiB is free. Including non-PyTorch memory, this process has 6.90 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 132.92 GiB is free. Including non-PyTorch memory, this process has 6.89 GiB memory in use. Of the allocated memory 5.11 GiB is allocated by PyTorch, and 302.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]:[W621 22:18:04.538881894 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:18:04.587446234 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:18:04.607482823 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:18:04.033039689 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:18:04.033062591 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:18:04.042937897 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:18:04.043044578 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:18:04.678004894 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:18:06.108000 817940 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 818011 closing signal SIGTERM +W0621 22:18:06.110000 817940 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 818012 closing signal SIGTERM +W0621 22:18:06.111000 817940 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 818014 closing signal SIGTERM +W0621 22:18:06.114000 817940 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 818015 closing signal SIGTERM +W0621 22:18:06.117000 2163941 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2164011 closing signal SIGTERM +W0621 22:18:06.115000 817940 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 818016 closing signal SIGTERM +W0621 22:18:06.117000 817940 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 818018 closing signal SIGTERM +W0621 22:18:06.120000 2163941 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2164012 closing signal SIGTERM +W0621 22:18:06.121000 2163941 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2164013 closing signal SIGTERM +W0621 22:18:06.124000 2163941 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2164014 closing signal SIGTERM +W0621 22:18:06.125000 2163941 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2164015 closing signal SIGTERM +W0621 22:18:06.127000 2163941 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2164017 closing signal SIGTERM +W0621 22:18:06.136000 817940 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 818019 closing signal SIGTERM +E0621 22:18:06.389000 2163941 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 5 (pid: 2164016) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +E0621 22:18:06.419000 817940 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 5 (pid: 818017) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:18:06 + host : fs-mbz-gpu-141 + rank : 7 (local_rank: 7) + exitcode : 1 (pid: 2164018) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:18:06 + host : fs-mbz-gpu-141 + rank : 5 (local_rank: 5) + exitcode : 1 (pid: 2164016) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:18:06 + host : fs-mbz-gpu-188 + rank : 13 (local_rank: 5) + exitcode : 1 (pid: 818017) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=131072 ++ PROF_CTX_LENGTH=131072 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L131072*tp2.cp8.bs4.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L131072*tp2.cp8.bs4.json' ']' ++ echo 'Running ctx_length=131072, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=4' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 131072 --max-position-embeddings 131072 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343239 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-188:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 131072 --max-position-embeddings 131072 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:18:10.940000 819750 site-packages/torch/distributed/run.py:766] +W0621 22:18:10.940000 819750 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:18:10.940000 819750 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:18:10.940000 819750 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:18:11.023000 2165802 site-packages/torch/distributed/run.py:766] +W0621 22:18:11.023000 2165802 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:18:11.023000 2165802 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:18:11.023000 2165802 site-packages/torch/distributed/run.py:766] ***************************************** +[rank2]:[W621 22:18:34.574904369 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:18:34.574905683 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank4]:[W621 22:18:34.575002286 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:18:34.003590071 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:18:34.003741537 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:18:34.003902536 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:18:34.003904004 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:18:34.003923600 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:18:34.004173852 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:18:34.583841064 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:18:34.004399903 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:18:34.584073363 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:18:34.584196277 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:18:34.591045696 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:18:34.099293355 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:18:34.716365719 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 131.40 GiB is free. Including non-PyTorch memory, this process has 8.40 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 131.42 GiB is free. Including non-PyTorch memory, this process has 8.39 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 131.42 GiB is free. Including non-PyTorch memory, this process has 8.39 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 131.42 GiB is free. Including non-PyTorch memory, this process has 8.39 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 131.42 GiB is free. Including non-PyTorch memory, this process has 8.39 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 131.42 GiB is free. Including non-PyTorch memory, this process has 8.39 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 131.42 GiB is free. Including non-PyTorch memory, this process has 8.39 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 131.42 GiB is free. Including non-PyTorch memory, this process has 8.39 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 131.40 GiB is free. Including non-PyTorch memory, this process has 8.40 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 131.40 GiB is free. Including non-PyTorch memory, this process has 8.40 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 131.40 GiB is free. Including non-PyTorch memory, this process has 8.40 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 131.40 GiB is free. Including non-PyTorch memory, this process has 8.40 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 131.40 GiB is free. Including non-PyTorch memory, this process has 8.40 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 131.40 GiB is free. Including non-PyTorch memory, this process has 8.40 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 131.42 GiB is free. Including non-PyTorch memory, this process has 8.39 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 131.40 GiB is free. Including non-PyTorch memory, this process has 8.40 GiB memory in use. Of the allocated memory 6.12 GiB is allocated by PyTorch, and 804.39 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]:[W621 22:18:49.043002346 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:18:49.043093886 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:18:49.043315646 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:18:49.651880020 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:18:49.113550084 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank1]:[W621 22:18:49.725911215 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:18:49.893933740 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:18:49.944408417 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:18:51.126000 2165802 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2165873 closing signal SIGTERM +W0621 22:18:51.128000 2165802 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2165874 closing signal SIGTERM +W0621 22:18:51.129000 2165802 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2165875 closing signal SIGTERM +W0621 22:18:51.132000 2165802 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2165876 closing signal SIGTERM +W0621 22:18:51.133000 2165802 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2165877 closing signal SIGTERM +W0621 22:18:51.135000 2165802 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2165879 closing signal SIGTERM +W0621 22:18:51.151000 2165802 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2165880 closing signal SIGTERM +W0621 22:18:51.161000 819750 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 819821 closing signal SIGTERM +W0621 22:18:51.164000 819750 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 819822 closing signal SIGTERM +W0621 22:18:51.165000 819750 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 819823 closing signal SIGTERM +W0621 22:18:51.168000 819750 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 819824 closing signal SIGTERM +W0621 22:18:51.169000 819750 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 819825 closing signal SIGTERM +W0621 22:18:51.172000 819750 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 819827 closing signal SIGTERM +E0621 22:18:51.493000 819750 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 5 (pid: 819826) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:18:51 + host : fs-mbz-gpu-188 + rank : 15 (local_rank: 7) + exitcode : 1 (pid: 819828) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:18:51 + host : fs-mbz-gpu-188 + rank : 13 (local_rank: 5) + exitcode : 1 (pid: 819826) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ +E0621 22:18:51.693000 2165802 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 5 (pid: 2165878) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +W0621 22:18:51.710000 2165802 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2165802_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +[W621 22:18:51.724715112 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-141]:52180, remote=[fs-mbz-gpu-188]:29500): Broken pipe +Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first): +frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string, std::allocator >) + 0x98 (0x14c7505785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so) +frame #1: + 0x5ba8afe (0x14c73985aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #2: + 0x5baa358 (0x14c73985c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #3: + 0x5babb3e (0x14c73985db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #4: c10d::TCPStore::doWait(c10::ArrayRef, std::allocator > >, std::chrono::duration >) + 0x1a6 (0x14c739857ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string, std::allocator > const&) + 0x33 (0x14c739857ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #6: c10d::TCPStore::get(std::__cxx11::basic_string, std::allocator > const&) + 0xab (0x14c739858f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #7: + 0xc0f526 (0x14c748b8b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) +frame #8: + 0x37f17d (0x14c7482fb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) + +frame #26: + 0x29d90 (0x14c75189cd90 in /lib/x86_64-linux-gnu/libc.so.6) +frame #27: __libc_start_main + 0x80 (0x14c75189ce40 in /lib/x86_64-linux-gnu/libc.so.6) + +W0621 22:18:51.725000 2165802 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2165802_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +[W621 22:18:51.738603863 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=3, addr=[fs-mbz-gpu-141]:52180, remote=[fs-mbz-gpu-188]:29500): Broken pipe +Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first): +frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string, std::allocator >) + 0x98 (0x14c7505785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so) +frame #1: + 0x5ba8afe (0x14c73985aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #2: + 0x5baa358 (0x14c73985c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #3: + 0x5babb3e (0x14c73985db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #4: c10d::TCPStore::doWait(c10::ArrayRef, std::allocator > >, std::chrono::duration >) + 0x1a6 (0x14c739857ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string, std::allocator > const&) + 0x33 (0x14c739857ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #6: c10d::TCPStore::get(std::__cxx11::basic_string, std::allocator > const&) + 0xab (0x14c739858f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so) +frame #7: + 0xc0f526 (0x14c748b8b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) +frame #8: + 0x37f17d (0x14c7482fb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so) + +frame #26: + 0x29d90 (0x14c75189cd90 in /lib/x86_64-linux-gnu/libc.so.6) +frame #27: __libc_start_main + 0x80 (0x14c75189ce40 in /lib/x86_64-linux-gnu/libc.so.6) + +W0621 22:18:51.738000 2165802 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-141_2165802_0' has failed to shutdown the rendezvous '343239' due to an error of type RendezvousConnectionError. +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:18:51 + host : fs-mbz-gpu-141 + rank : 5 (local_rank: 5) + exitcode : 1 (pid: 2165878) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ set +x