Upload train.sh with huggingface_hub
Browse files
train.sh
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/bin/bash
|
| 2 |
+
NUM_PROC=$1
|
| 3 |
+
shift
|
| 4 |
+
# the localhost fixes are from https://github.com/pytorch/pytorch/issues/73320
|
| 5 |
+
torchrun --rdzv_backend c10d --rdzv_endpoint localhost:0 --nproc_per_node=$NUM_PROC train.py "$@"
|