GindaChen commited on
Commit
28276c9
·
verified ·
1 Parent(s): cb745f8

Upload folder using huggingface_hub

Browse files
attnserver.run_attnserver.slurm.sh.343188.out.log CHANGED
@@ -124818,3 +124818,633 @@ batch tensor after cp: position_ids torch.Size([1, 16384])
124818
  Start exporting trace 8
124819
  Done exporting trace 8
124820
  [2025-06-21 21:18:21] iteration 9/ 10 | consumed samples: 9 | elapsed time per iteration (ms): 128291.4 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 16777216.0 | number of skipped iterations: 1 | number of nan iterations: 0 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
124818
  Start exporting trace 8
124819
  Done exporting trace 8
124820
  [2025-06-21 21:18:21] iteration 9/ 10 | consumed samples: 9 | elapsed time per iteration (ms): 128291.4 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 16777216.0 | number of skipped iterations: 1 | number of nan iterations: 0 |
124821
+ batch tensor: tokens torch.Size([1, 131072])
124822
+ batch tensor: labels torch.Size([1, 131072])
124823
+ batch tensor: loss_mask torch.Size([1, 131072])
124824
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124825
+ batch tensor: position_ids torch.Size([1, 131072])
124826
+ batch tensor after cp: tokens torch.Size([1, 16384])
124827
+ batch tensor after cp: labels torch.Size([1, 16384])
124828
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124829
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124830
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124831
+ batch tensor: tokens torch.Size([1, 131072])
124832
+ batch tensor: labels torch.Size([1, 131072])
124833
+ batch tensor: loss_mask torch.Size([1, 131072])
124834
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124835
+ batch tensor: position_ids torch.Size([1, 131072])
124836
+ batch tensor after cp: tokens torch.Size([1, 16384])
124837
+ batch tensor after cp: labels torch.Size([1, 16384])
124838
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124839
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124840
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124841
+ batch tensor: tokens torch.Size([1, 131072])
124842
+ batch tensor: labels torch.Size([1, 131072])
124843
+ batch tensor: loss_mask torch.Size([1, 131072])
124844
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124845
+ batch tensor: position_ids torch.Size([1, 131072])
124846
+ batch tensor after cp: tokens torch.Size([1, 16384])
124847
+ batch tensor after cp: labels torch.Size([1, 16384])
124848
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124849
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124850
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124851
+ batch tensor: tokens torch.Size([1, 131072])
124852
+ batch tensor: labels torch.Size([1, 131072])
124853
+ batch tensor: loss_mask torch.Size([1, 131072])
124854
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124855
+ batch tensor: position_ids torch.Size([1, 131072])
124856
+ batch tensor after cp: tokens torch.Size([1, 16384])
124857
+ batch tensor after cp: labels torch.Size([1, 16384])
124858
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124859
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124860
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124861
+ batch tensor: tokens torch.Size([1, 131072])
124862
+ batch tensor: labels torch.Size([1, 131072])
124863
+ batch tensor: loss_mask torch.Size([1, 131072])
124864
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124865
+ batch tensor: position_ids torch.Size([1, 131072])
124866
+ batch tensor after cp: tokens torch.Size([1, 16384])
124867
+ batch tensor after cp: labels torch.Size([1, 16384])
124868
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124869
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124870
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124871
+ batch tensor: tokens torch.Size([1, 131072])
124872
+ batch tensor: labels torch.Size([1, 131072])
124873
+ batch tensor: loss_mask torch.Size([1, 131072])
124874
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124875
+ batch tensor: position_ids torch.Size([1, 131072])
124876
+ batch tensor after cp: tokens torch.Size([1, 16384])
124877
+ batch tensor after cp: labels torch.Size([1, 16384])
124878
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124879
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124880
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124881
+ batch tensor: tokens torch.Size([1, 131072])
124882
+ batch tensor: labels torch.Size([1, 131072])
124883
+ batch tensor: loss_mask torch.Size([1, 131072])
124884
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124885
+ batch tensor: position_ids torch.Size([1, 131072])
124886
+ batch tensor after cp: tokens torch.Size([1, 16384])
124887
+ batch tensor after cp: labels torch.Size([1, 16384])
124888
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124889
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124890
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124891
+ batch tensor: tokens torch.Size([1, 131072])
124892
+ batch tensor: labels torch.Size([1, 131072])
124893
+ batch tensor: loss_mask torch.Size([1, 131072])
124894
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124895
+ batch tensor: position_ids torch.Size([1, 131072])
124896
+ batch tensor after cp: tokens torch.Size([1, 16384])
124897
+ batch tensor after cp: labels torch.Size([1, 16384])
124898
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124899
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124900
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124901
+ batch tensor: tokens torch.Size([1, 131072])
124902
+ batch tensor: labels torch.Size([1, 131072])
124903
+ batch tensor: loss_mask torch.Size([1, 131072])
124904
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124905
+ batch tensor: position_ids torch.Size([1, 131072])
124906
+ batch tensor after cp: tokens torch.Size([1, 16384])
124907
+ batch tensor after cp: labels torch.Size([1, 16384])
124908
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124909
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124910
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124911
+ batch tensor: tokens torch.Size([1, 131072])
124912
+ batch tensor: labels torch.Size([1, 131072])
124913
+ batch tensor: loss_mask torch.Size([1, 131072])
124914
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124915
+ batch tensor: position_ids torch.Size([1, 131072])
124916
+ batch tensor after cp: tokens torch.Size([1, 16384])
124917
+ batch tensor after cp: labels torch.Size([1, 16384])
124918
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124919
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124920
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124921
+ batch tensor: tokens torch.Size([1, 131072])
124922
+ batch tensor: labels torch.Size([1, 131072])
124923
+ batch tensor: loss_mask torch.Size([1, 131072])
124924
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124925
+ batch tensor: position_ids torch.Size([1, 131072])
124926
+ batch tensor after cp: tokens torch.Size([1, 16384])
124927
+ batch tensor after cp: labels torch.Size([1, 16384])
124928
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124929
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124930
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124931
+ batch tensor: tokens torch.Size([1, 131072])
124932
+ batch tensor: labels torch.Size([1, 131072])
124933
+ batch tensor: loss_mask torch.Size([1, 131072])
124934
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124935
+ batch tensor: position_ids torch.Size([1, 131072])
124936
+ batch tensor after cp: tokens torch.Size([1, 16384])
124937
+ batch tensor after cp: labels torch.Size([1, 16384])
124938
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124939
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124940
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124941
+ batch tensor: tokens torch.Size([1, 131072])
124942
+ batch tensor: labels torch.Size([1, 131072])
124943
+ batch tensor: loss_mask torch.Size([1, 131072])
124944
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124945
+ batch tensor: position_ids torch.Size([1, 131072])
124946
+ batch tensor after cp: tokens torch.Size([1, 16384])
124947
+ batch tensor after cp: labels torch.Size([1, 16384])
124948
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124949
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124950
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124951
+ batch tensor: tokens torch.Size([1, 131072])
124952
+ batch tensor: labels torch.Size([1, 131072])
124953
+ batch tensor: loss_mask torch.Size([1, 131072])
124954
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124955
+ batch tensor: position_ids torch.Size([1, 131072])
124956
+ batch tensor after cp: tokens torch.Size([1, 16384])
124957
+ batch tensor after cp: labels torch.Size([1, 16384])
124958
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124959
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124960
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124961
+ batch tensor: tokens torch.Size([1, 131072])
124962
+ batch tensor: labels torch.Size([1, 131072])
124963
+ batch tensor: loss_mask torch.Size([1, 131072])
124964
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124965
+ batch tensor: position_ids torch.Size([1, 131072])
124966
+ batch tensor after cp: tokens torch.Size([1, 16384])
124967
+ batch tensor after cp: labels torch.Size([1, 16384])
124968
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124969
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124970
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124971
+ batch tensor: tokens torch.Size([1, 131072])
124972
+ batch tensor: labels torch.Size([1, 131072])
124973
+ batch tensor: loss_mask torch.Size([1, 131072])
124974
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124975
+ batch tensor: position_ids torch.Size([1, 131072])
124976
+ batch tensor after cp: tokens torch.Size([1, 16384])
124977
+ batch tensor after cp: labels torch.Size([1, 16384])
124978
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124979
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124980
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124981
+ batch tensor: tokens torch.Size([1, 131072])
124982
+ batch tensor: labels torch.Size([1, 131072])
124983
+ batch tensor: loss_mask torch.Size([1, 131072])
124984
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124985
+ batch tensor: position_ids torch.Size([1, 131072])
124986
+ batch tensor after cp: tokens torch.Size([1, 16384])
124987
+ batch tensor after cp: labels torch.Size([1, 16384])
124988
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124989
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
124990
+ batch tensor after cp: position_ids torch.Size([1, 16384])
124991
+ batch tensor: tokens torch.Size([1, 131072])
124992
+ batch tensor: labels torch.Size([1, 131072])
124993
+ batch tensor: loss_mask torch.Size([1, 131072])
124994
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
124995
+ batch tensor: position_ids torch.Size([1, 131072])
124996
+ batch tensor after cp: tokens torch.Size([1, 16384])
124997
+ batch tensor after cp: labels torch.Size([1, 16384])
124998
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
124999
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125000
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125001
+ batch tensor: tokens torch.Size([1, 131072])
125002
+ batch tensor: labels torch.Size([1, 131072])
125003
+ batch tensor: loss_mask torch.Size([1, 131072])
125004
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125005
+ batch tensor: position_ids torch.Size([1, 131072])
125006
+ batch tensor after cp: tokens torch.Size([1, 16384])
125007
+ batch tensor after cp: labels torch.Size([1, 16384])
125008
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125009
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125010
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125011
+ batch tensor: tokens torch.Size([1, 131072])
125012
+ batch tensor: labels torch.Size([1, 131072])
125013
+ batch tensor: loss_mask torch.Size([1, 131072])
125014
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125015
+ batch tensor: position_ids torch.Size([1, 131072])
125016
+ batch tensor after cp: tokens torch.Size([1, 16384])
125017
+ batch tensor after cp: labels torch.Size([1, 16384])
125018
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125019
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125020
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125021
+ batch tensor: tokens torch.Size([1, 131072])
125022
+ batch tensor: labels torch.Size([1, 131072])
125023
+ batch tensor: loss_mask torch.Size([1, 131072])
125024
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125025
+ batch tensor: position_ids torch.Size([1, 131072])
125026
+ batch tensor after cp: tokens torch.Size([1, 16384])
125027
+ batch tensor after cp: labels torch.Size([1, 16384])
125028
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125029
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125030
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125031
+ batch tensor: tokens torch.Size([1, 131072])
125032
+ batch tensor: labels torch.Size([1, 131072])
125033
+ batch tensor: loss_mask torch.Size([1, 131072])
125034
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125035
+ batch tensor: position_ids torch.Size([1, 131072])
125036
+ batch tensor after cp: tokens torch.Size([1, 16384])
125037
+ batch tensor after cp: labels torch.Size([1, 16384])
125038
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125039
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125040
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125041
+ batch tensor: tokens torch.Size([1, 131072])
125042
+ batch tensor: labels torch.Size([1, 131072])
125043
+ batch tensor: loss_mask torch.Size([1, 131072])
125044
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125045
+ batch tensor: position_ids torch.Size([1, 131072])
125046
+ batch tensor after cp: tokens torch.Size([1, 16384])
125047
+ batch tensor after cp: labels torch.Size([1, 16384])
125048
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125049
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125050
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125051
+ batch tensor: tokens torch.Size([1, 131072])
125052
+ batch tensor: labels torch.Size([1, 131072])
125053
+ batch tensor: loss_mask torch.Size([1, 131072])
125054
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125055
+ batch tensor: position_ids torch.Size([1, 131072])
125056
+ batch tensor after cp: tokens torch.Size([1, 16384])
125057
+ batch tensor after cp: labels torch.Size([1, 16384])
125058
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125059
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125060
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125061
+ batch tensor: tokens torch.Size([1, 131072])
125062
+ batch tensor: labels torch.Size([1, 131072])
125063
+ batch tensor: loss_mask torch.Size([1, 131072])
125064
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125065
+ batch tensor: position_ids torch.Size([1, 131072])
125066
+ batch tensor after cp: tokens torch.Size([1, 16384])
125067
+ batch tensor after cp: labels torch.Size([1, 16384])
125068
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125069
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125070
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125071
+ batch tensor: tokens torch.Size([1, 131072])
125072
+ batch tensor: labels torch.Size([1, 131072])
125073
+ batch tensor: loss_mask torch.Size([1, 131072])
125074
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125075
+ batch tensor: position_ids torch.Size([1, 131072])
125076
+ batch tensor after cp: tokens torch.Size([1, 16384])
125077
+ batch tensor after cp: labels torch.Size([1, 16384])
125078
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125079
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125080
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125081
+ batch tensor: tokens torch.Size([1, 131072])
125082
+ batch tensor: labels torch.Size([1, 131072])
125083
+ batch tensor: loss_mask torch.Size([1, 131072])
125084
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125085
+ batch tensor: position_ids torch.Size([1, 131072])
125086
+ batch tensor after cp: tokens torch.Size([1, 16384])
125087
+ batch tensor after cp: labels torch.Size([1, 16384])
125088
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125089
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125090
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125091
+ batch tensor: tokens torch.Size([1, 131072])
125092
+ batch tensor: labels torch.Size([1, 131072])
125093
+ batch tensor: loss_mask torch.Size([1, 131072])
125094
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125095
+ batch tensor: position_ids torch.Size([1, 131072])
125096
+ batch tensor after cp: tokens torch.Size([1, 16384])
125097
+ batch tensor after cp: labels torch.Size([1, 16384])
125098
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125099
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125100
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125101
+ batch tensor: tokens torch.Size([1, 131072])
125102
+ batch tensor: labels torch.Size([1, 131072])
125103
+ batch tensor: loss_mask torch.Size([1, 131072])
125104
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125105
+ batch tensor: position_ids torch.Size([1, 131072])
125106
+ batch tensor after cp: tokens torch.Size([1, 16384])
125107
+ batch tensor after cp: labels torch.Size([1, 16384])
125108
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125109
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125110
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125111
+ batch tensor: tokens torch.Size([1, 131072])
125112
+ batch tensor: labels torch.Size([1, 131072])
125113
+ batch tensor: loss_mask torch.Size([1, 131072])
125114
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125115
+ batch tensor: position_ids torch.Size([1, 131072])
125116
+ batch tensor after cp: tokens torch.Size([1, 16384])
125117
+ batch tensor after cp: labels torch.Size([1, 16384])
125118
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125119
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125120
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125121
+ batch tensor: tokens torch.Size([1, 131072])
125122
+ batch tensor: labels torch.Size([1, 131072])
125123
+ batch tensor: loss_mask torch.Size([1, 131072])
125124
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125125
+ batch tensor: position_ids torch.Size([1, 131072])
125126
+ batch tensor after cp: tokens torch.Size([1, 16384])
125127
+ batch tensor after cp: labels torch.Size([1, 16384])
125128
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125129
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125130
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125131
+ batch tensor: tokens torch.Size([1, 131072])
125132
+ batch tensor: labels torch.Size([1, 131072])
125133
+ batch tensor: loss_mask torch.Size([1, 131072])
125134
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125135
+ batch tensor: position_ids torch.Size([1, 131072])
125136
+ batch tensor after cp: tokens torch.Size([1, 16384])
125137
+ batch tensor after cp: labels torch.Size([1, 16384])
125138
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125139
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125140
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125141
+ batch tensor: tokens torch.Size([1, 131072])
125142
+ batch tensor: labels torch.Size([1, 131072])
125143
+ batch tensor: loss_mask torch.Size([1, 131072])
125144
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125145
+ batch tensor: position_ids torch.Size([1, 131072])
125146
+ batch tensor after cp: tokens torch.Size([1, 16384])
125147
+ batch tensor after cp: labels torch.Size([1, 16384])
125148
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125149
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125150
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125151
+ batch tensor: tokens torch.Size([1, 131072])
125152
+ batch tensor: labels torch.Size([1, 131072])
125153
+ batch tensor: loss_mask torch.Size([1, 131072])
125154
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125155
+ batch tensor: position_ids torch.Size([1, 131072])
125156
+ batch tensor after cp: tokens torch.Size([1, 16384])
125157
+ batch tensor after cp: labels torch.Size([1, 16384])
125158
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125159
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125160
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125161
+ batch tensor: tokens torch.Size([1, 131072])
125162
+ batch tensor: labels torch.Size([1, 131072])
125163
+ batch tensor: loss_mask torch.Size([1, 131072])
125164
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125165
+ batch tensor: position_ids torch.Size([1, 131072])
125166
+ batch tensor after cp: tokens torch.Size([1, 16384])
125167
+ batch tensor after cp: labels torch.Size([1, 16384])
125168
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125169
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125170
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125171
+ batch tensor: tokens torch.Size([1, 131072])
125172
+ batch tensor: labels torch.Size([1, 131072])
125173
+ batch tensor: loss_mask torch.Size([1, 131072])
125174
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125175
+ batch tensor: position_ids torch.Size([1, 131072])
125176
+ batch tensor after cp: tokens torch.Size([1, 16384])
125177
+ batch tensor after cp: labels torch.Size([1, 16384])
125178
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125179
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125180
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125181
+ batch tensor: tokens torch.Size([1, 131072])
125182
+ batch tensor: labels torch.Size([1, 131072])
125183
+ batch tensor: loss_mask torch.Size([1, 131072])
125184
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125185
+ batch tensor: position_ids torch.Size([1, 131072])
125186
+ batch tensor after cp: tokens torch.Size([1, 16384])
125187
+ batch tensor after cp: labels torch.Size([1, 16384])
125188
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125189
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125190
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125191
+ batch tensor: tokens torch.Size([1, 131072])
125192
+ batch tensor: labels torch.Size([1, 131072])
125193
+ batch tensor: loss_mask torch.Size([1, 131072])
125194
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125195
+ batch tensor: position_ids torch.Size([1, 131072])
125196
+ batch tensor after cp: tokens torch.Size([1, 16384])
125197
+ batch tensor after cp: labels torch.Size([1, 16384])
125198
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125199
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125200
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125201
+ batch tensor: tokens torch.Size([1, 131072])
125202
+ batch tensor: labels torch.Size([1, 131072])
125203
+ batch tensor: loss_mask torch.Size([1, 131072])
125204
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125205
+ batch tensor: position_ids torch.Size([1, 131072])
125206
+ batch tensor after cp: tokens torch.Size([1, 16384])
125207
+ batch tensor after cp: labels torch.Size([1, 16384])
125208
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125209
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125210
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125211
+ batch tensor: tokens torch.Size([1, 131072])
125212
+ batch tensor: labels torch.Size([1, 131072])
125213
+ batch tensor: loss_mask torch.Size([1, 131072])
125214
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125215
+ batch tensor: position_ids torch.Size([1, 131072])
125216
+ batch tensor after cp: tokens torch.Size([1, 16384])
125217
+ batch tensor after cp: labels torch.Size([1, 16384])
125218
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125219
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125220
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125221
+ batch tensor: tokens torch.Size([1, 131072])
125222
+ batch tensor: labels torch.Size([1, 131072])
125223
+ batch tensor: loss_mask torch.Size([1, 131072])
125224
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125225
+ batch tensor: position_ids torch.Size([1, 131072])
125226
+ batch tensor after cp: tokens torch.Size([1, 16384])
125227
+ batch tensor after cp: labels torch.Size([1, 16384])
125228
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125229
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125230
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125231
+ batch tensor: tokens torch.Size([1, 131072])
125232
+ batch tensor: labels torch.Size([1, 131072])
125233
+ batch tensor: loss_mask torch.Size([1, 131072])
125234
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125235
+ batch tensor: position_ids torch.Size([1, 131072])
125236
+ batch tensor after cp: tokens torch.Size([1, 16384])
125237
+ batch tensor after cp: labels torch.Size([1, 16384])
125238
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125239
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125240
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125241
+ batch tensor: tokens torch.Size([1, 131072])
125242
+ batch tensor: labels torch.Size([1, 131072])
125243
+ batch tensor: loss_mask torch.Size([1, 131072])
125244
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125245
+ batch tensor: position_ids torch.Size([1, 131072])
125246
+ batch tensor after cp: tokens torch.Size([1, 16384])
125247
+ batch tensor after cp: labels torch.Size([1, 16384])
125248
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125249
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125250
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125251
+ batch tensor: tokens torch.Size([1, 131072])
125252
+ batch tensor: labels torch.Size([1, 131072])
125253
+ batch tensor: loss_mask torch.Size([1, 131072])
125254
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125255
+ batch tensor: position_ids torch.Size([1, 131072])
125256
+ batch tensor after cp: tokens torch.Size([1, 16384])
125257
+ batch tensor after cp: labels torch.Size([1, 16384])
125258
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125259
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125260
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125261
+ batch tensor: tokens torch.Size([1, 131072])
125262
+ batch tensor: labels torch.Size([1, 131072])
125263
+ batch tensor: loss_mask torch.Size([1, 131072])
125264
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125265
+ batch tensor: position_ids torch.Size([1, 131072])
125266
+ batch tensor after cp: tokens torch.Size([1, 16384])
125267
+ batch tensor after cp: labels torch.Size([1, 16384])
125268
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125269
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125270
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125271
+ batch tensor: tokens torch.Size([1, 131072])
125272
+ batch tensor: labels torch.Size([1, 131072])
125273
+ batch tensor: loss_mask torch.Size([1, 131072])
125274
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125275
+ batch tensor: position_ids torch.Size([1, 131072])
125276
+ batch tensor after cp: tokens torch.Size([1, 16384])
125277
+ batch tensor after cp: labels torch.Size([1, 16384])
125278
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125279
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125280
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125281
+ batch tensor: tokens torch.Size([1, 131072])
125282
+ batch tensor: labels torch.Size([1, 131072])
125283
+ batch tensor: loss_mask torch.Size([1, 131072])
125284
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125285
+ batch tensor: position_ids torch.Size([1, 131072])
125286
+ batch tensor after cp: tokens torch.Size([1, 16384])
125287
+ batch tensor after cp: labels torch.Size([1, 16384])
125288
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125289
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125290
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125291
+ batch tensor: tokens torch.Size([1, 131072])
125292
+ batch tensor: labels torch.Size([1, 131072])
125293
+ batch tensor: loss_mask torch.Size([1, 131072])
125294
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125295
+ batch tensor: position_ids torch.Size([1, 131072])
125296
+ batch tensor after cp: tokens torch.Size([1, 16384])
125297
+ batch tensor after cp: labels torch.Size([1, 16384])
125298
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125299
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125300
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125301
+ batch tensor: tokens torch.Size([1, 131072])
125302
+ batch tensor: labels torch.Size([1, 131072])
125303
+ batch tensor: loss_mask torch.Size([1, 131072])
125304
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125305
+ batch tensor: position_ids torch.Size([1, 131072])
125306
+ batch tensor after cp: tokens torch.Size([1, 16384])
125307
+ batch tensor after cp: labels torch.Size([1, 16384])
125308
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125309
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125310
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125311
+ batch tensor: tokens torch.Size([1, 131072])
125312
+ batch tensor: labels torch.Size([1, 131072])
125313
+ batch tensor: loss_mask torch.Size([1, 131072])
125314
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125315
+ batch tensor: position_ids torch.Size([1, 131072])
125316
+ batch tensor after cp: tokens torch.Size([1, 16384])
125317
+ batch tensor after cp: labels torch.Size([1, 16384])
125318
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125319
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125320
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125321
+ batch tensor: tokens torch.Size([1, 131072])
125322
+ batch tensor: labels torch.Size([1, 131072])
125323
+ batch tensor: loss_mask torch.Size([1, 131072])
125324
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125325
+ batch tensor: position_ids torch.Size([1, 131072])
125326
+ batch tensor after cp: tokens torch.Size([1, 16384])
125327
+ batch tensor after cp: labels torch.Size([1, 16384])
125328
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125329
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125330
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125331
+ batch tensor: tokens torch.Size([1, 131072])
125332
+ batch tensor: labels torch.Size([1, 131072])
125333
+ batch tensor: loss_mask torch.Size([1, 131072])
125334
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125335
+ batch tensor: position_ids torch.Size([1, 131072])
125336
+ batch tensor after cp: tokens torch.Size([1, 16384])
125337
+ batch tensor after cp: labels torch.Size([1, 16384])
125338
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125339
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125340
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125341
+ batch tensor: tokens torch.Size([1, 131072])
125342
+ batch tensor: labels torch.Size([1, 131072])
125343
+ batch tensor: loss_mask torch.Size([1, 131072])
125344
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125345
+ batch tensor: position_ids torch.Size([1, 131072])
125346
+ batch tensor after cp: tokens torch.Size([1, 16384])
125347
+ batch tensor after cp: labels torch.Size([1, 16384])
125348
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125349
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125350
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125351
+ batch tensor: tokens torch.Size([1, 131072])
125352
+ batch tensor: labels torch.Size([1, 131072])
125353
+ batch tensor: loss_mask torch.Size([1, 131072])
125354
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125355
+ batch tensor: position_ids torch.Size([1, 131072])
125356
+ batch tensor after cp: tokens torch.Size([1, 16384])
125357
+ batch tensor after cp: labels torch.Size([1, 16384])
125358
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125359
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125360
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125361
+ batch tensor: tokens torch.Size([1, 131072])
125362
+ batch tensor: labels torch.Size([1, 131072])
125363
+ batch tensor: loss_mask torch.Size([1, 131072])
125364
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125365
+ batch tensor: position_ids torch.Size([1, 131072])
125366
+ batch tensor after cp: tokens torch.Size([1, 16384])
125367
+ batch tensor after cp: labels torch.Size([1, 16384])
125368
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125369
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125370
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125371
+ batch tensor: tokens torch.Size([1, 131072])
125372
+ batch tensor: labels torch.Size([1, 131072])
125373
+ batch tensor: loss_mask torch.Size([1, 131072])
125374
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125375
+ batch tensor: position_ids torch.Size([1, 131072])
125376
+ batch tensor after cp: tokens torch.Size([1, 16384])
125377
+ batch tensor after cp: labels torch.Size([1, 16384])
125378
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125379
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125380
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125381
+ batch tensor: tokens torch.Size([1, 131072])
125382
+ batch tensor: labels torch.Size([1, 131072])
125383
+ batch tensor: loss_mask torch.Size([1, 131072])
125384
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125385
+ batch tensor: position_ids torch.Size([1, 131072])
125386
+ batch tensor after cp: tokens torch.Size([1, 16384])
125387
+ batch tensor after cp: labels torch.Size([1, 16384])
125388
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125389
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125390
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125391
+ batch tensor: tokens torch.Size([1, 131072])
125392
+ batch tensor: labels torch.Size([1, 131072])
125393
+ batch tensor: loss_mask torch.Size([1, 131072])
125394
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125395
+ batch tensor: position_ids torch.Size([1, 131072])
125396
+ batch tensor after cp: tokens torch.Size([1, 16384])
125397
+ batch tensor after cp: labels torch.Size([1, 16384])
125398
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125399
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125400
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125401
+ batch tensor: tokens torch.Size([1, 131072])
125402
+ batch tensor: labels torch.Size([1, 131072])
125403
+ batch tensor: loss_mask torch.Size([1, 131072])
125404
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125405
+ batch tensor: position_ids torch.Size([1, 131072])
125406
+ batch tensor after cp: tokens torch.Size([1, 16384])
125407
+ batch tensor after cp: labels torch.Size([1, 16384])
125408
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125409
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125410
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125411
+ batch tensor: tokens torch.Size([1, 131072])
125412
+ batch tensor: labels torch.Size([1, 131072])
125413
+ batch tensor: loss_mask torch.Size([1, 131072])
125414
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125415
+ batch tensor: position_ids torch.Size([1, 131072])
125416
+ batch tensor after cp: tokens torch.Size([1, 16384])
125417
+ batch tensor after cp: labels torch.Size([1, 16384])
125418
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125419
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125420
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125421
+ batch tensor: tokens torch.Size([1, 131072])
125422
+ batch tensor: labels torch.Size([1, 131072])
125423
+ batch tensor: loss_mask torch.Size([1, 131072])
125424
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125425
+ batch tensor: position_ids torch.Size([1, 131072])
125426
+ batch tensor after cp: tokens torch.Size([1, 16384])
125427
+ batch tensor after cp: labels torch.Size([1, 16384])
125428
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125429
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125430
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125431
+ batch tensor: tokens torch.Size([1, 131072])
125432
+ batch tensor: labels torch.Size([1, 131072])
125433
+ batch tensor: loss_mask torch.Size([1, 131072])
125434
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125435
+ batch tensor: position_ids torch.Size([1, 131072])
125436
+ batch tensor after cp: tokens torch.Size([1, 16384])
125437
+ batch tensor after cp: labels torch.Size([1, 16384])
125438
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125439
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125440
+ batch tensor after cp: position_ids torch.Size([1, 16384])
125441
+ batch tensor: tokens torch.Size([1, 131072])
125442
+ batch tensor: labels torch.Size([1, 131072])
125443
+ batch tensor: loss_mask torch.Size([1, 131072])
125444
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
125445
+ batch tensor: position_ids torch.Size([1, 131072])
125446
+ batch tensor after cp: tokens torch.Size([1, 16384])
125447
+ batch tensor after cp: labels torch.Size([1, 16384])
125448
+ batch tensor after cp: loss_mask torch.Size([1, 16384])
125449
+ batch tensor after cp: attention_mask torch.Size([1, 1, 16384, 131072])
125450
+ batch tensor after cp: position_ids torch.Size([1, 16384])
attnserver.run_attnserver.slurm.sh.343195.out.log CHANGED
@@ -67763,3 +67763,293 @@ batch tensor after cp: position_ids torch.Size([1, 32768])
67763
  Start exporting trace 6
67764
  Done exporting trace 6
67765
  [2025-06-21 21:18:27] iteration 7/ 10 | consumed samples: 7 | elapsed time per iteration (ms): 152588.3 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 67108864.0 | number of skipped iterations: 1 | number of nan iterations: 0 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
67763
  Start exporting trace 6
67764
  Done exporting trace 6
67765
  [2025-06-21 21:18:27] iteration 7/ 10 | consumed samples: 7 | elapsed time per iteration (ms): 152588.3 | learning rate: 0.000000E+00 | global batch size: 1 | loss scale: 67108864.0 | number of skipped iterations: 1 | number of nan iterations: 0 |
67766
+ batch tensor: tokens torch.Size([1, 131072])
67767
+ batch tensor: labels torch.Size([1, 131072])
67768
+ batch tensor: loss_mask torch.Size([1, 131072])
67769
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67770
+ batch tensor: position_ids torch.Size([1, 131072])
67771
+ batch tensor after cp: tokens torch.Size([1, 32768])
67772
+ batch tensor after cp: labels torch.Size([1, 32768])
67773
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67774
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67775
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67776
+ batch tensor: tokens torch.Size([1, 131072])
67777
+ batch tensor: labels torch.Size([1, 131072])
67778
+ batch tensor: loss_mask torch.Size([1, 131072])
67779
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67780
+ batch tensor: position_ids torch.Size([1, 131072])
67781
+ batch tensor after cp: tokens torch.Size([1, 32768])
67782
+ batch tensor after cp: labels torch.Size([1, 32768])
67783
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67784
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67785
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67786
+ batch tensor: tokens torch.Size([1, 131072])
67787
+ batch tensor: labels torch.Size([1, 131072])
67788
+ batch tensor: loss_mask torch.Size([1, 131072])
67789
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67790
+ batch tensor: position_ids torch.Size([1, 131072])
67791
+ batch tensor after cp: tokens torch.Size([1, 32768])
67792
+ batch tensor after cp: labels torch.Size([1, 32768])
67793
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67794
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67795
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67796
+ batch tensor: tokens torch.Size([1, 131072])
67797
+ batch tensor: labels torch.Size([1, 131072])
67798
+ batch tensor: loss_mask torch.Size([1, 131072])
67799
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67800
+ batch tensor: position_ids torch.Size([1, 131072])
67801
+ batch tensor after cp: tokens torch.Size([1, 32768])
67802
+ batch tensor after cp: labels torch.Size([1, 32768])
67803
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67804
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67805
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67806
+ batch tensor: tokens torch.Size([1, 131072])
67807
+ batch tensor: labels torch.Size([1, 131072])
67808
+ batch tensor: loss_mask torch.Size([1, 131072])
67809
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67810
+ batch tensor: position_ids torch.Size([1, 131072])
67811
+ batch tensor after cp: tokens torch.Size([1, 32768])
67812
+ batch tensor after cp: labels torch.Size([1, 32768])
67813
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67814
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67815
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67816
+ batch tensor: tokens torch.Size([1, 131072])
67817
+ batch tensor: labels torch.Size([1, 131072])
67818
+ batch tensor: loss_mask torch.Size([1, 131072])
67819
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67820
+ batch tensor: position_ids torch.Size([1, 131072])
67821
+ batch tensor after cp: tokens torch.Size([1, 32768])
67822
+ batch tensor after cp: labels torch.Size([1, 32768])
67823
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67824
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67825
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67826
+ batch tensor: tokens torch.Size([1, 131072])
67827
+ batch tensor: labels torch.Size([1, 131072])
67828
+ batch tensor: loss_mask torch.Size([1, 131072])
67829
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67830
+ batch tensor: position_ids torch.Size([1, 131072])
67831
+ batch tensor after cp: tokens torch.Size([1, 32768])
67832
+ batch tensor after cp: labels torch.Size([1, 32768])
67833
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67834
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67835
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67836
+ batch tensor: tokens torch.Size([1, 131072])
67837
+ batch tensor: labels torch.Size([1, 131072])
67838
+ batch tensor: loss_mask torch.Size([1, 131072])
67839
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67840
+ batch tensor: position_ids torch.Size([1, 131072])
67841
+ batch tensor after cp: tokens torch.Size([1, 32768])
67842
+ batch tensor after cp: labels torch.Size([1, 32768])
67843
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67844
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67845
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67846
+ batch tensor: tokens torch.Size([1, 131072])
67847
+ batch tensor: labels torch.Size([1, 131072])
67848
+ batch tensor: loss_mask torch.Size([1, 131072])
67849
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67850
+ batch tensor: position_ids torch.Size([1, 131072])
67851
+ batch tensor after cp: tokens torch.Size([1, 32768])
67852
+ batch tensor after cp: labels torch.Size([1, 32768])
67853
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67854
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67855
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67856
+ batch tensor: tokens torch.Size([1, 131072])
67857
+ batch tensor: labels torch.Size([1, 131072])
67858
+ batch tensor: loss_mask torch.Size([1, 131072])
67859
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67860
+ batch tensor: position_ids torch.Size([1, 131072])
67861
+ batch tensor after cp: tokens torch.Size([1, 32768])
67862
+ batch tensor after cp: labels torch.Size([1, 32768])
67863
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67864
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67865
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67866
+ batch tensor: tokens torch.Size([1, 131072])
67867
+ batch tensor: labels torch.Size([1, 131072])
67868
+ batch tensor: loss_mask torch.Size([1, 131072])
67869
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67870
+ batch tensor: position_ids torch.Size([1, 131072])
67871
+ batch tensor after cp: tokens torch.Size([1, 32768])
67872
+ batch tensor after cp: labels torch.Size([1, 32768])
67873
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67874
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67875
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67876
+ batch tensor: tokens torch.Size([1, 131072])
67877
+ batch tensor: labels torch.Size([1, 131072])
67878
+ batch tensor: loss_mask torch.Size([1, 131072])
67879
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67880
+ batch tensor: position_ids torch.Size([1, 131072])
67881
+ batch tensor after cp: tokens torch.Size([1, 32768])
67882
+ batch tensor after cp: labels torch.Size([1, 32768])
67883
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67884
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67885
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67886
+ batch tensor: tokens torch.Size([1, 131072])
67887
+ batch tensor: labels torch.Size([1, 131072])
67888
+ batch tensor: loss_mask torch.Size([1, 131072])
67889
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67890
+ batch tensor: position_ids torch.Size([1, 131072])
67891
+ batch tensor after cp: tokens torch.Size([1, 32768])
67892
+ batch tensor after cp: labels torch.Size([1, 32768])
67893
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67894
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67895
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67896
+ batch tensor: tokens torch.Size([1, 131072])
67897
+ batch tensor: labels torch.Size([1, 131072])
67898
+ batch tensor: loss_mask torch.Size([1, 131072])
67899
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67900
+ batch tensor: position_ids torch.Size([1, 131072])
67901
+ batch tensor after cp: tokens torch.Size([1, 32768])
67902
+ batch tensor after cp: labels torch.Size([1, 32768])
67903
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67904
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67905
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67906
+ batch tensor: tokens torch.Size([1, 131072])
67907
+ batch tensor: labels torch.Size([1, 131072])
67908
+ batch tensor: loss_mask torch.Size([1, 131072])
67909
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67910
+ batch tensor: position_ids torch.Size([1, 131072])
67911
+ batch tensor after cp: tokens torch.Size([1, 32768])
67912
+ batch tensor after cp: labels torch.Size([1, 32768])
67913
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67914
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67915
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67916
+ batch tensor: tokens torch.Size([1, 131072])
67917
+ batch tensor: labels torch.Size([1, 131072])
67918
+ batch tensor: loss_mask torch.Size([1, 131072])
67919
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67920
+ batch tensor: position_ids torch.Size([1, 131072])
67921
+ batch tensor after cp: tokens torch.Size([1, 32768])
67922
+ batch tensor after cp: labels torch.Size([1, 32768])
67923
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67924
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67925
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67926
+ batch tensor: tokens torch.Size([1, 131072])
67927
+ batch tensor: labels torch.Size([1, 131072])
67928
+ batch tensor: loss_mask torch.Size([1, 131072])
67929
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67930
+ batch tensor: position_ids torch.Size([1, 131072])
67931
+ batch tensor: tokens torch.Size([1, 131072])
67932
+ batch tensor: labels torch.Size([1, 131072])
67933
+ batch tensor: loss_mask torch.Size([1, 131072])
67934
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67935
+ batch tensor: position_ids torch.Size([1, 131072])
67936
+ batch tensor after cp: tokens torch.Size([1, 32768])
67937
+ batch tensor after cp: tokens torch.Size([1, 32768])
67938
+ batch tensor after cp: labels torch.Size([1, 32768])
67939
+ batch tensor after cp: labels torch.Size([1, 32768])
67940
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67941
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67942
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67943
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67944
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67945
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67946
+ batch tensor: tokens torch.Size([1, 131072])
67947
+ batch tensor: labels torch.Size([1, 131072])
67948
+ batch tensor: loss_mask torch.Size([1, 131072])
67949
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67950
+ batch tensor: position_ids torch.Size([1, 131072])
67951
+ batch tensor after cp: tokens torch.Size([1, 32768])
67952
+ batch tensor after cp: labels torch.Size([1, 32768])
67953
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67954
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67955
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67956
+ batch tensor: tokens torch.Size([1, 131072])
67957
+ batch tensor: labels torch.Size([1, 131072])
67958
+ batch tensor: loss_mask torch.Size([1, 131072])
67959
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67960
+ batch tensor: position_ids torch.Size([1, 131072])
67961
+ batch tensor after cp: tokens torch.Size([1, 32768])
67962
+ batch tensor after cp: labels torch.Size([1, 32768])
67963
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67964
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67965
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67966
+ batch tensor: tokens torch.Size([1, 131072])
67967
+ batch tensor: labels torch.Size([1, 131072])
67968
+ batch tensor: loss_mask torch.Size([1, 131072])
67969
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67970
+ batch tensor: position_ids torch.Size([1, 131072])
67971
+ batch tensor after cp: tokens torch.Size([1, 32768])
67972
+ batch tensor after cp: labels torch.Size([1, 32768])
67973
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67974
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67975
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67976
+ batch tensor: tokens torch.Size([1, 131072])
67977
+ batch tensor: labels torch.Size([1, 131072])
67978
+ batch tensor: loss_mask torch.Size([1, 131072])
67979
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67980
+ batch tensor: position_ids torch.Size([1, 131072])
67981
+ batch tensor after cp: tokens torch.Size([1, 32768])
67982
+ batch tensor after cp: labels torch.Size([1, 32768])
67983
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67984
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67985
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67986
+ batch tensor: tokens torch.Size([1, 131072])
67987
+ batch tensor: labels torch.Size([1, 131072])
67988
+ batch tensor: loss_mask torch.Size([1, 131072])
67989
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
67990
+ batch tensor: position_ids torch.Size([1, 131072])
67991
+ batch tensor after cp: tokens torch.Size([1, 32768])
67992
+ batch tensor after cp: labels torch.Size([1, 32768])
67993
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
67994
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
67995
+ batch tensor after cp: position_ids torch.Size([1, 32768])
67996
+ batch tensor: tokens torch.Size([1, 131072])
67997
+ batch tensor: labels torch.Size([1, 131072])
67998
+ batch tensor: loss_mask torch.Size([1, 131072])
67999
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
68000
+ batch tensor: position_ids torch.Size([1, 131072])
68001
+ batch tensor after cp: tokens torch.Size([1, 32768])
68002
+ batch tensor after cp: labels torch.Size([1, 32768])
68003
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
68004
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
68005
+ batch tensor after cp: position_ids torch.Size([1, 32768])
68006
+ batch tensor: tokens torch.Size([1, 131072])
68007
+ batch tensor: labels torch.Size([1, 131072])
68008
+ batch tensor: loss_mask torch.Size([1, 131072])
68009
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
68010
+ batch tensor: position_ids torch.Size([1, 131072])
68011
+ batch tensor after cp: tokens torch.Size([1, 32768])
68012
+ batch tensor after cp: labels torch.Size([1, 32768])
68013
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
68014
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
68015
+ batch tensor after cp: position_ids torch.Size([1, 32768])
68016
+ batch tensor: tokens torch.Size([1, 131072])
68017
+ batch tensor: labels torch.Size([1, 131072])
68018
+ batch tensor: loss_mask torch.Size([1, 131072])
68019
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
68020
+ batch tensor: position_ids torch.Size([1, 131072])
68021
+ batch tensor after cp: tokens torch.Size([1, 32768])
68022
+ batch tensor after cp: labels torch.Size([1, 32768])
68023
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
68024
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
68025
+ batch tensor after cp: position_ids torch.Size([1, 32768])
68026
+ batch tensor: tokens torch.Size([1, 131072])
68027
+ batch tensor: labels torch.Size([1, 131072])
68028
+ batch tensor: loss_mask torch.Size([1, 131072])
68029
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
68030
+ batch tensor: position_ids torch.Size([1, 131072])
68031
+ batch tensor after cp: tokens torch.Size([1, 32768])
68032
+ batch tensor after cp: labels torch.Size([1, 32768])
68033
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
68034
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
68035
+ batch tensor after cp: position_ids torch.Size([1, 32768])
68036
+ batch tensor: tokens torch.Size([1, 131072])
68037
+ batch tensor: labels torch.Size([1, 131072])
68038
+ batch tensor: loss_mask torch.Size([1, 131072])
68039
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
68040
+ batch tensor: position_ids torch.Size([1, 131072])
68041
+ batch tensor after cp: tokens torch.Size([1, 32768])
68042
+ batch tensor after cp: labels torch.Size([1, 32768])
68043
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
68044
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
68045
+ batch tensor after cp: position_ids torch.Size([1, 32768])
68046
+ batch tensor: tokens torch.Size([1, 131072])
68047
+ batch tensor: labels torch.Size([1, 131072])
68048
+ batch tensor: loss_mask torch.Size([1, 131072])
68049
+ batch tensor: attention_mask torch.Size([1, 1, 131072, 131072])
68050
+ batch tensor: position_ids torch.Size([1, 131072])
68051
+ batch tensor after cp: tokens torch.Size([1, 32768])
68052
+ batch tensor after cp: labels torch.Size([1, 32768])
68053
+ batch tensor after cp: loss_mask torch.Size([1, 32768])
68054
+ batch tensor after cp: attention_mask torch.Size([1, 1, 32768, 131072])
68055
+ batch tensor after cp: position_ids torch.Size([1, 32768])
attnserver.run_attnserver.slurm.sh.343196.out.log CHANGED
The diff for this file is too large to render. See raw diff
 
attnserver.run_attnserver.slurm.sh.343200.err.log CHANGED
The diff for this file is too large to render. See raw diff
 
attnserver.run_attnserver.slurm.sh.343200.out.log CHANGED
The diff for this file is too large to render. See raw diff
 
attnserver.run_attnserver.slurm.sh.343202.err.log CHANGED
@@ -6782,3 +6782,310 @@ W0621 21:13:47.545000 3922086 site-packages/torch/distributed/run.py:766] ******
6782
  [rank0]: return io.open(self, mode, buffering, encoding, errors, newline)
6783
  [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6784
  [rank0]: FileNotFoundError: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/.metadata.tmp'
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6782
  [rank0]: return io.open(self, mode, buffering, encoding, errors, newline)
6783
  [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6784
  [rank0]: FileNotFoundError: [Errno 2] No such file or directory: 'gpt-checkpoint/iter_0000010/.metadata.tmp'
6785
+ [rank0]:[W621 21:19:04.979946813 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())
6786
+ W0621 21:19:13.223000 3922086 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3922159 closing signal SIGTERM
6787
+ W0621 21:19:13.227000 3922086 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3922160 closing signal SIGTERM
6788
+ W0621 21:19:13.230000 3922086 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3922161 closing signal SIGTERM
6789
+ W0621 21:19:13.231000 3922086 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3922162 closing signal SIGTERM
6790
+ W0621 21:19:13.236000 3922086 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3922163 closing signal SIGTERM
6791
+ W0621 21:19:13.240000 3922086 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3922164 closing signal SIGTERM
6792
+ W0621 21:19:13.260000 3922086 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 3922165 closing signal SIGTERM
6793
+ E0621 21:19:15.487000 3922086 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 0 (pid: 3922158) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
6794
+ Traceback (most recent call last):
6795
+ File "<frozen runpy>", line 198, in _run_module_as_main
6796
+ File "<frozen runpy>", line 88, in _run_code
6797
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
6798
+ main()
6799
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
6800
+ return arg(*args, **kwargs)
6801
+ ^^^^^^^^^^^^^^^^^^^^
6802
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
6803
+ launch(args)
6804
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
6805
+ run(args)
6806
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
6807
+ elastic_launch(
6808
+ 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__
6809
+ return launch_agent(self._config, self._entrypoint, list(args))
6810
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6811
+ 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
6812
+ raise ChildFailedError(
6813
+ torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
6814
+ ============================================================
6815
+ ./pretrain_gpt_profile.py FAILED
6816
+ ------------------------------------------------------------
6817
+ Failures:
6818
+ <NO_OTHER_FAILURES>
6819
+ ------------------------------------------------------------
6820
+ Root Cause (first observed failure):
6821
+ [0]:
6822
+ time : 2025-06-21_21:19:13
6823
+ host : fs-mbz-gpu-728
6824
+ rank : 0 (local_rank: 0)
6825
+ exitcode : 1 (pid: 3922158)
6826
+ error_file: <N/A>
6827
+ traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
6828
+ ============================================================
6829
+ + set +x
6830
+ [rank14]:[W621 21:19:16.597385554 TCPStore.cpp:125] [c10d] recvValue failed on SocketImpl(fd=95, addr=[fs-mbz-gpu-865]:57156, remote=[fs-mbz-gpu-728]:37481): failed to recv, got 0 bytes
6831
+ Exception raised from recvBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:678 (most recent call first):
6832
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x15088b9785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6833
+ frame #1: <unknown function> + 0x5ba8afe (0x15087485aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6834
+ frame #2: <unknown function> + 0x5baae40 (0x15087485ce40 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6835
+ frame #3: <unknown function> + 0x5bab74a (0x15087485d74a in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6836
+ frame #4: c10d::TCPStore::check(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&) + 0x2a9 (0x1508748571a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6837
+ frame #5: c10d::ProcessGroupNCCL::heartbeatMonitor() + 0x379 (0x150831a509a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cuda.so)
6838
+ frame #6: <unknown function> + 0xd3b6d (0x15088b4f1b6d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/../lib/libstdc++.so.6)
6839
+ frame #7: <unknown function> + 0x94ac3 (0x15088ca36ac3 in /lib/x86_64-linux-gnu/libc.so.6)
6840
+ frame #8: <unknown function> + 0x126850 (0x15088cac8850 in /lib/x86_64-linux-gnu/libc.so.6)
6841
+
6842
+ [rank15]:[W621 21:19:16.597385455 TCPStore.cpp:125] [c10d] recvValue failed on SocketImpl(fd=95, addr=[fs-mbz-gpu-865]:57224, remote=[fs-mbz-gpu-728]:37481): failed to recv, got 0 bytes
6843
+ Exception raised from recvBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:678 (most recent call first):
6844
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x145f943785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6845
+ frame #1: <unknown function> + 0x5ba8afe (0x145f7d25aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6846
+ frame #2: <unknown function> + 0x5baae40 (0x145f7d25ce40 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6847
+ frame #3: <unknown function> + 0x5bab74a (0x145f7d25d74a in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6848
+ frame #4: c10d::TCPStore::check(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&) + 0x2a9 (0x145f7d2571a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6849
+ frame #5: c10d::ProcessGroupNCCL::heartbeatMonitor() + 0x379 (0x145f3a4509a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cuda.so)
6850
+ frame #6: <unknown function> + 0xd3b6d (0x145f93ef1b6d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/../lib/libstdc++.so.6)
6851
+ frame #7: <unknown function> + 0x94ac3 (0x145f953b8ac3 in /lib/x86_64-linux-gnu/libc.so.6)
6852
+ frame #8: <unknown function> + 0x126850 (0x145f9544a850 in /lib/x86_64-linux-gnu/libc.so.6)
6853
+
6854
+ [rank10]:[W621 21:19:16.597522826 TCPStore.cpp:125] [c10d] recvValue failed on SocketImpl(fd=95, addr=[fs-mbz-gpu-865]:57206, remote=[fs-mbz-gpu-728]:37481): failed to recv, got 0 bytes
6855
+ Exception raised from recvBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:678 (most recent call first):
6856
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14d75bd785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6857
+ frame #1: <unknown function> + 0x5ba8afe (0x14d744c5aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6858
+ frame #2: <unknown function> + 0x5baae40 (0x14d744c5ce40 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6859
+ frame #3: <unknown function> + 0x5bab74a (0x14d744c5d74a in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6860
+ frame #4: c10d::TCPStore::check(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&) + 0x2a9 (0x14d744c571a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6861
+ frame #5: c10d::ProcessGroupNCCL::heartbeatMonitor() + 0x379 (0x14d701e509a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cuda.so)
6862
+ frame #6: <unknown function> + 0xd3b6d (0x14d75b8f1b6d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/../lib/libstdc++.so.6)
6863
+ frame #7: <unknown function> + 0x94ac3 (0x14d75cdc5ac3 in /lib/x86_64-linux-gnu/libc.so.6)
6864
+ frame #8: <unknown function> + 0x126850 (0x14d75ce57850 in /lib/x86_64-linux-gnu/libc.so.6)
6865
+
6866
+ [rank14]:[W621 21:19:16.601790151 ProcessGroupNCCL.cpp:1659] [PG ID 0 PG GUID 0(default_pg) Rank 14] Failed to check the "should dump" flag on TCPStore, (maybe TCPStore server has shut down too early), with error: failed to recv, got 0 bytes
6867
+ [rank15]:[W621 21:19:16.601801332 ProcessGroupNCCL.cpp:1659] [PG ID 0 PG GUID 0(default_pg) Rank 15] Failed to check the "should dump" flag on TCPStore, (maybe TCPStore server has shut down too early), with error: failed to recv, got 0 bytes
6868
+ [rank10]:[W621 21:19:16.601904060 ProcessGroupNCCL.cpp:1659] [PG ID 0 PG GUID 0(default_pg) Rank 10] Failed to check the "should dump" flag on TCPStore, (maybe TCPStore server has shut down too early), with error: failed to recv, got 0 bytes
6869
+ [rank8]:[W621 21:19:16.629327826 TCPStore.cpp:125] [c10d] recvValue failed on SocketImpl(fd=75, addr=[fs-mbz-gpu-865]:57148, remote=[fs-mbz-gpu-728]:37481): failed to recv, got 0 bytes
6870
+ Exception raised from recvBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:678 (most recent call first):
6871
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x1465d09785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6872
+ frame #1: <unknown function> + 0x5ba8afe (0x1465b985aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6873
+ frame #2: <unknown function> + 0x5baae40 (0x1465b985ce40 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6874
+ frame #3: <unknown function> + 0x5bab74a (0x1465b985d74a in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6875
+ frame #4: c10d::TCPStore::check(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&) + 0x2a9 (0x1465b98571a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6876
+ frame #5: c10d::ProcessGroupNCCL::heartbeatMonitor() + 0x379 (0x146576a509a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cuda.so)
6877
+ frame #6: <unknown function> + 0xd3b6d (0x1465d04f1b6d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/../lib/libstdc++.so.6)
6878
+ frame #7: <unknown function> + 0x94ac3 (0x1465d1a14ac3 in /lib/x86_64-linux-gnu/libc.so.6)
6879
+ frame #8: <unknown function> + 0x126850 (0x1465d1aa6850 in /lib/x86_64-linux-gnu/libc.so.6)
6880
+
6881
+ [rank8]:[W621 21:19:16.633647516 ProcessGroupNCCL.cpp:1659] [PG ID 0 PG GUID 0(default_pg) Rank 8] Failed to check the "should dump" flag on TCPStore, (maybe TCPStore server has shut down too early), with error: failed to recv, got 0 bytes
6882
+ [rank11]:[W621 21:19:16.688718409 TCPStore.cpp:125] [c10d] recvValue failed on SocketImpl(fd=95, addr=[fs-mbz-gpu-865]:57220, remote=[fs-mbz-gpu-728]:37481): failed to recv, got 0 bytes
6883
+ Exception raised from recvBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:678 (most recent call first):
6884
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x146437b785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6885
+ frame #1: <unknown function> + 0x5ba8afe (0x146420e5aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6886
+ frame #2: <unknown function> + 0x5baae40 (0x146420e5ce40 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6887
+ frame #3: <unknown function> + 0x5bab74a (0x146420e5d74a in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6888
+ frame #4: c10d::TCPStore::check(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&) + 0x2a9 (0x146420e571a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6889
+ frame #5: c10d::ProcessGroupNCCL::heartbeatMonitor() + 0x379 (0x1463de0509a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cuda.so)
6890
+ frame #6: <unknown function> + 0xd3b6d (0x1463ce019b6d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/../lib/libstdc++.so.6)
6891
+ frame #7: <unknown function> + 0x94ac3 (0x146438f6dac3 in /lib/x86_64-linux-gnu/libc.so.6)
6892
+ frame #8: <unknown function> + 0x126850 (0x146438fff850 in /lib/x86_64-linux-gnu/libc.so.6)
6893
+
6894
+ [rank11]:[W621 21:19:16.692730257 ProcessGroupNCCL.cpp:1659] [PG ID 0 PG GUID 0(default_pg) Rank 11] Failed to check the "should dump" flag on TCPStore, (maybe TCPStore server has shut down too early), with error: failed to recv, got 0 bytes
6895
+ [rank12]:[W621 21:19:16.688662135 TCPStore.cpp:125] [c10d] recvValue failed on SocketImpl(fd=95, addr=[fs-mbz-gpu-865]:57166, remote=[fs-mbz-gpu-728]:37481): failed to recv, got 0 bytes
6896
+ Exception raised from recvBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:678 (most recent call first):
6897
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14776f3785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6898
+ frame #1: <unknown function> + 0x5ba8afe (0x14775865aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6899
+ frame #2: <unknown function> + 0x5baae40 (0x14775865ce40 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6900
+ frame #3: <unknown function> + 0x5bab74a (0x14775865d74a in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6901
+ frame #4: c10d::TCPStore::check(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&) + 0x2a9 (0x1477586571a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6902
+ frame #5: c10d::ProcessGroupNCCL::heartbeatMonitor() + 0x379 (0x1477158509a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cuda.so)
6903
+ frame #6: <unknown function> + 0xd3b6d (0x147705819b6d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/../lib/libstdc++.so.6)
6904
+ frame #7: <unknown function> + 0x94ac3 (0x147770710ac3 in /lib/x86_64-linux-gnu/libc.so.6)
6905
+ frame #8: <unknown function> + 0x126850 (0x1477707a2850 in /lib/x86_64-linux-gnu/libc.so.6)
6906
+
6907
+ [rank9]:[W621 21:19:16.688664327 TCPStore.cpp:125] [c10d] recvValue failed on SocketImpl(fd=95, addr=[fs-mbz-gpu-865]:57180, remote=[fs-mbz-gpu-728]:37481): failed to recv, got 0 bytes
6908
+ Exception raised from recvBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:678 (most recent call first):
6909
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x14d335b785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6910
+ frame #1: <unknown function> + 0x5ba8afe (0x14d31ea5aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6911
+ frame #2: <unknown function> + 0x5baae40 (0x14d31ea5ce40 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6912
+ frame #3: <unknown function> + 0x5bab74a (0x14d31ea5d74a in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6913
+ frame #4: c10d::TCPStore::check(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&) + 0x2a9 (0x14d31ea571a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6914
+ frame #5: c10d::ProcessGroupNCCL::heartbeatMonitor() + 0x379 (0x14d2dbc509a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cuda.so)
6915
+ frame #6: <unknown function> + 0xd3b6d (0x14d3356f1b6d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/../lib/libstdc++.so.6)
6916
+ frame #7: <unknown function> + 0x94ac3 (0x14d336be8ac3 in /lib/x86_64-linux-gnu/libc.so.6)
6917
+ frame #8: <unknown function> + 0x126850 (0x14d336c7a850 in /lib/x86_64-linux-gnu/libc.so.6)
6918
+
6919
+ [rank13]:[W621 21:19:16.688672768 TCPStore.cpp:125] [c10d] recvValue failed on SocketImpl(fd=95, addr=[fs-mbz-gpu-865]:57196, remote=[fs-mbz-gpu-728]:37481): failed to recv, got 0 bytes
6920
+ Exception raised from recvBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:678 (most recent call first):
6921
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x1512e83785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6922
+ frame #1: <unknown function> + 0x5ba8afe (0x1512d165aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6923
+ frame #2: <unknown function> + 0x5baae40 (0x1512d165ce40 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6924
+ frame #3: <unknown function> + 0x5bab74a (0x1512d165d74a in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6925
+ frame #4: c10d::TCPStore::check(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&) + 0x2a9 (0x1512d16571a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6926
+ frame #5: c10d::ProcessGroupNCCL::heartbeatMonitor() + 0x379 (0x15128e8509a9 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cuda.so)
6927
+ frame #6: <unknown function> + 0xd3b6d (0x15127e819b6d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/../lib/libstdc++.so.6)
6928
+ frame #7: <unknown function> + 0x94ac3 (0x1512e96b9ac3 in /lib/x86_64-linux-gnu/libc.so.6)
6929
+ frame #8: <unknown function> + 0x126850 (0x1512e974b850 in /lib/x86_64-linux-gnu/libc.so.6)
6930
+
6931
+ [rank12]:[W621 21:19:16.692870336 ProcessGroupNCCL.cpp:1659] [PG ID 0 PG GUID 0(default_pg) Rank 12] Failed to check the "should dump" flag on TCPStore, (maybe TCPStore server has shut down too early), with error: failed to recv, got 0 bytes
6932
+ [rank9]:[W621 21:19:16.692884070 ProcessGroupNCCL.cpp:1659] [PG ID 0 PG GUID 0(default_pg) Rank 9] Failed to check the "should dump" flag on TCPStore, (maybe TCPStore server has shut down too early), with error: failed to recv, got 0 bytes
6933
+ [rank13]:[W621 21:19:16.692933680 ProcessGroupNCCL.cpp:1659] [PG ID 0 PG GUID 0(default_pg) Rank 13] Failed to check the "should dump" flag on TCPStore, (maybe TCPStore server has shut down too early), with error: failed to recv, got 0 bytes
6934
+ W0621 21:19:16.473000 2522629 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2522699 closing signal SIGTERM
6935
+ W0621 21:19:16.478000 2522629 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2522700 closing signal SIGTERM
6936
+ W0621 21:19:16.480000 2522629 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2522701 closing signal SIGTERM
6937
+ W0621 21:19:16.485000 2522629 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2522702 closing signal SIGTERM
6938
+ W0621 21:19:16.487000 2522629 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2522703 closing signal SIGTERM
6939
+ W0621 21:19:16.489000 2522629 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2522704 closing signal SIGTERM
6940
+ W0621 21:19:16.533000 2522629 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2522705 closing signal SIGTERM
6941
+ W0621 21:19:16.548000 2522629 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2522707 closing signal SIGTERM
6942
+ [W621 21:19:18.748504106 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-865]:42814, remote=[fs-mbz-gpu-728]:29500): Broken pipe
6943
+ Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
6944
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x1474ef7785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6945
+ frame #1: <unknown function> + 0x5ba8afe (0x1474d865aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6946
+ frame #2: <unknown function> + 0x5baa358 (0x1474d865c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6947
+ frame #3: <unknown function> + 0x5babb3e (0x1474d865db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6948
+ frame #4: c10d::TCPStore::doWait(c10::ArrayRef<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::chrono::duration<long, std::ratio<1l, 1000l> >) + 0x1a6 (0x1474d8657ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6949
+ frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x33 (0x1474d8657ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6950
+ frame #6: c10d::TCPStore::get(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xab (0x1474d8658f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6951
+ frame #7: <unknown function> + 0xc0f526 (0x1474e798b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
6952
+ frame #8: <unknown function> + 0x37f17d (0x1474e70fb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
6953
+ <omitting python frames>
6954
+ frame #17: <unknown function> + 0x94ac3 (0x1474f07f3ac3 in /lib/x86_64-linux-gnu/libc.so.6)
6955
+ frame #18: <unknown function> + 0x126850 (0x1474f0885850 in /lib/x86_64-linux-gnu/libc.so.6)
6956
+
6957
+ W0621 21:19:18.428000 2522629 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1341] The node 'fs-mbz-gpu-865_2522629_0' has failed to send a keep-alive heartbeat to the rendezvous '343202' due to an error of type RendezvousConnectionError.
6958
+ [W621 21:19:23.757992050 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-865]:42814, remote=[fs-mbz-gpu-728]:29500): Broken pipe
6959
+ Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
6960
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x1474ef7785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6961
+ frame #1: <unknown function> + 0x5ba8afe (0x1474d865aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6962
+ frame #2: <unknown function> + 0x5baa358 (0x1474d865c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6963
+ frame #3: <unknown function> + 0x5babb3e (0x1474d865db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6964
+ frame #4: c10d::TCPStore::doWait(c10::ArrayRef<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::chrono::duration<long, std::ratio<1l, 1000l> >) + 0x1a6 (0x1474d8657ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6965
+ frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x33 (0x1474d8657ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6966
+ frame #6: c10d::TCPStore::get(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xab (0x1474d8658f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6967
+ frame #7: <unknown function> + 0xc0f526 (0x1474e798b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
6968
+ frame #8: <unknown function> + 0x37f17d (0x1474e70fb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
6969
+ <omitting python frames>
6970
+ frame #17: <unknown function> + 0x94ac3 (0x1474f07f3ac3 in /lib/x86_64-linux-gnu/libc.so.6)
6971
+ frame #18: <unknown function> + 0x126850 (0x1474f0885850 in /lib/x86_64-linux-gnu/libc.so.6)
6972
+
6973
+ W0621 21:19:23.435000 2522629 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1341] The node 'fs-mbz-gpu-865_2522629_0' has failed to send a keep-alive heartbeat to the rendezvous '343202' due to an error of type RendezvousConnectionError.
6974
+ [W621 21:19:23.821313515 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-865]:42814, remote=[fs-mbz-gpu-728]:29500): Broken pipe
6975
+ Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
6976
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x1474ef7785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6977
+ frame #1: <unknown function> + 0x5ba8afe (0x1474d865aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6978
+ frame #2: <unknown function> + 0x5baa358 (0x1474d865c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6979
+ frame #3: <unknown function> + 0x5babb3e (0x1474d865db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6980
+ frame #4: c10d::TCPStore::doWait(c10::ArrayRef<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::chrono::duration<long, std::ratio<1l, 1000l> >) + 0x1a6 (0x1474d8657ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6981
+ frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x33 (0x1474d8657ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6982
+ frame #6: c10d::TCPStore::get(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xab (0x1474d8658f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6983
+ frame #7: <unknown function> + 0xc0f526 (0x1474e798b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
6984
+ frame #8: <unknown function> + 0x37f17d (0x1474e70fb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
6985
+ <omitting python frames>
6986
+ frame #26: <unknown function> + 0x29d90 (0x1474f0788d90 in /lib/x86_64-linux-gnu/libc.so.6)
6987
+ frame #27: __libc_start_main + 0x80 (0x1474f0788e40 in /lib/x86_64-linux-gnu/libc.so.6)
6988
+
6989
+ W0621 21:19:23.505000 2522629 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-865_2522629_0' has failed to shutdown the rendezvous '343202' due to an error of type RendezvousConnectionError.
6990
+ [W621 21:19:23.835472402 TCPStore.cpp:106] [c10d] sendBytes failed on SocketImpl(fd=4, addr=[fs-mbz-gpu-865]:42814, remote=[fs-mbz-gpu-728]:29500): Broken pipe
6991
+ Exception raised from sendBytes at /pytorch/torch/csrc/distributed/c10d/Utils.hpp:653 (most recent call first):
6992
+ frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x1474ef7785e8 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libc10.so)
6993
+ frame #1: <unknown function> + 0x5ba8afe (0x1474d865aafe in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6994
+ frame #2: <unknown function> + 0x5baa358 (0x1474d865c358 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6995
+ frame #3: <unknown function> + 0x5babb3e (0x1474d865db3e in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6996
+ frame #4: c10d::TCPStore::doWait(c10::ArrayRef<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::chrono::duration<long, std::ratio<1l, 1000l> >) + 0x1a6 (0x1474d8657ac6 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6997
+ frame #5: c10d::TCPStore::doGet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x33 (0x1474d8657ea3 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6998
+ frame #6: c10d::TCPStore::get(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xab (0x1474d8658f8b in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
6999
+ frame #7: <unknown function> + 0xc0f526 (0x1474e798b526 in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
7000
+ frame #8: <unknown function> + 0x37f17d (0x1474e70fb17d in /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
7001
+ <omitting python frames>
7002
+ frame #26: <unknown function> + 0x29d90 (0x1474f0788d90 in /lib/x86_64-linux-gnu/libc.so.6)
7003
+ frame #27: __libc_start_main + 0x80 (0x1474f0788e40 in /lib/x86_64-linux-gnu/libc.so.6)
7004
+
7005
+ W0621 21:19:23.516000 2522629 site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1292] The node 'fs-mbz-gpu-865_2522629_0' has failed to shutdown the rendezvous '343202' due to an error of type RendezvousConnectionError.
7006
+ Traceback (most recent call last):
7007
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 117, in _call_store
7008
+ return getattr(self._store, store_op)(*args, **kwargs)
7009
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7010
+ torch.distributed.DistNetworkError: failed to recv, got 0 bytes
7011
+
7012
+ The above exception was the direct cause of the following exception:
7013
+
7014
+ Traceback (most recent call last):
7015
+ File "<frozen runpy>", line 198, in _run_module_as_main
7016
+ File "<frozen runpy>", line 88, in _run_code
7017
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in <module>
7018
+ main()
7019
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper
7020
+ return arg(*args, **kwargs)
7021
+ ^^^^^^^^^^^^^^^^^^^^
7022
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main
7023
+ launch(args)
7024
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch
7025
+ run(args)
7026
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
7027
+ elastic_launch(
7028
+ 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__
7029
+ return launch_agent(self._config, self._entrypoint, list(args))
7030
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7031
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
7032
+ result = agent.run()
7033
+ ^^^^^^^^^^^
7034
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
7035
+ result = f(*args, **kwargs)
7036
+ ^^^^^^^^^^^^^^^^^^
7037
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
7038
+ result = self._invoke_run(role)
7039
+ ^^^^^^^^^^^^^^^^^^^^^^
7040
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/agent/server/api.py", line 906, in _invoke_run
7041
+ num_nodes_waiting = rdzv_handler.num_nodes_waiting()
7042
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7043
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py", line 1263, in num_nodes_waiting
7044
+ self._state_holder.sync()
7045
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py", line 437, in sync
7046
+ get_response = self._backend.get_state()
7047
+ ^^^^^^^^^^^^^^^^^^^^^^^^^
7048
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 75, in get_state
7049
+ base64_state: bytes = self._call_store("get", self._key)
7050
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7051
+ File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 119, in _call_store
7052
+ raise RendezvousConnectionError(
7053
+ torch.distributed.elastic.rendezvous.api.RendezvousConnectionError: The connection to the C10d store has failed. See inner exception for details.
7054
+ + set +x
7055
+ + for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
7056
+ + export PROF_CTX_LENGTH=81920
7057
+ + PROF_CTX_LENGTH=81920
7058
+ + name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L81920*tp8.cp2.bs2.json'
7059
+ + '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L81920*tp8.cp2.bs2.json' ']'
7060
+ + echo 'Running ctx_length=81920, TP_SIZE=8, CP_SIZE=2, BATCH_SIZE=2'
7061
+ + srun bash ./attnserver.sh
7062
+ + which python3
7063
+ + python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343202 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-728:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 8 --context-parallel-size 2 --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/
7064
+ + which python3
7065
+ + python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343202 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-728:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 8 --context-parallel-size 2 --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/
7066
+ /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
7067
+ and will be removed in future. Use torchrun.
7068
+ Note that --use-env is set by default in torchrun.
7069
+ If your script expects `--local-rank` argument to be set, please
7070
+ change it to read from `os.environ['LOCAL_RANK']` instead. See
7071
+ https://pytorch.org/docs/stable/distributed.html#launch-utility for
7072
+ further instructions
7073
+
7074
+ main()
7075
+ W0621 21:19:26.938000 2525849 site-packages/torch/distributed/run.py:766]
7076
+ W0621 21:19:26.938000 2525849 site-packages/torch/distributed/run.py:766] *****************************************
7077
+ W0621 21:19:26.938000 2525849 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.
7078
+ W0621 21:19:26.938000 2525849 site-packages/torch/distributed/run.py:766] *****************************************
7079
+ /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
7080
+ and will be removed in future. Use torchrun.
7081
+ Note that --use-env is set by default in torchrun.
7082
+ If your script expects `--local-rank` argument to be set, please
7083
+ change it to read from `os.environ['LOCAL_RANK']` instead. See
7084
+ https://pytorch.org/docs/stable/distributed.html#launch-utility for
7085
+ further instructions
7086
+
7087
+ main()
7088
+ W0621 21:19:27.143000 3925433 site-packages/torch/distributed/run.py:766]
7089
+ W0621 21:19:27.143000 3925433 site-packages/torch/distributed/run.py:766] *****************************************
7090
+ W0621 21:19:27.143000 3925433 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.
7091
+ W0621 21:19:27.143000 3925433 site-packages/torch/distributed/run.py:766] *****************************************
attnserver.run_attnserver.slurm.sh.343202.out.log CHANGED
@@ -29921,3 +29921,22 @@ DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750540712.95
29921
  DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750540712.9507585, 15, gather: 1.8799049854278564
29922
  DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750540712.9546258, 0, gather: 0.007202863693237305
29923
  DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.0099s
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
29921
  DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750540712.9507585, 15, gather: 1.8799049854278564
29922
  DEBUG:megatron.core.dist_checkpointing.strategies.state_dict_saver:1750540712.9546258, 0, gather: 0.007202863693237305
29923
  DEBUG:megatron.core.dist_checkpointing.strategies.async_utils:finalize took 0.0099s
29924
+ Running ctx_length=81920, TP_SIZE=8, CP_SIZE=2, BATCH_SIZE=2
29925
+ Cleaning up checkpoint directory: gpt-checkpoint
29926
+ Cleaning up checkpoint directory: gpt-checkpoint
29927
+ --------------------------------
29928
+ CTX_LENGTH: 81920
29929
+ TP_SIZE: 8
29930
+ CP_SIZE: 2
29931
+ CHECKPOINT_PATH: gpt-checkpoint
29932
+ PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron
29933
+ --------------------------------
29934
+ /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
29935
+ --------------------------------
29936
+ CTX_LENGTH: 81920
29937
+ TP_SIZE: 8
29938
+ CP_SIZE: 2
29939
+ CHECKPOINT_PATH: gpt-checkpoint
29940
+ PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron
29941
+ --------------------------------
29942
+ /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
attnserver.run_attnserver.slurm.sh.343203.err.log CHANGED
@@ -695,3 +695,104 @@ W0621 21:18:19.257000 758676 site-packages/torch/distributed/run.py:766]
695
  W0621 21:18:19.257000 758676 site-packages/torch/distributed/run.py:766] *****************************************
696
  W0621 21:18:19.257000 758676 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.
697
  W0621 21:18:19.257000 758676 site-packages/torch/distributed/run.py:766] *****************************************
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
695
  W0621 21:18:19.257000 758676 site-packages/torch/distributed/run.py:766] *****************************************
696
  W0621 21:18:19.257000 758676 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.
697
  W0621 21:18:19.257000 758676 site-packages/torch/distributed/run.py:766] *****************************************
698
+ [rank5]:[W621 21:18:42.008152355 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.
699
+ [rank13]:[W621 21:18:42.669951592 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.
700
+ [rank1]:[W621 21:18:42.009315982 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.
701
+ [rank9]:[W621 21:18:42.671191991 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.
702
+ [rank15]:[W621 21:18:42.671262833 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.
703
+ [rank12]:[W621 21:18:42.671775299 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.
704
+ [rank7]:[W621 21:18:42.012603791 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.
705
+ [rank4]:[W621 21:18:42.014142683 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.
706
+ [rank8]:[W621 21:18:43.829344186 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.
707
+ [rank0]:[W621 21:18:44.272155162 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.
708
+ [rank14]:[W621 21:18:44.950915815 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.
709
+ [rank11]:[W621 21:18:44.951789519 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.
710
+ [rank10]:[W621 21:18:44.952640951 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.
711
+ [rank3]:[W621 21:18:44.294979689 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.
712
+ [rank2]:[W621 21:18:44.295832026 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.
713
+ [rank6]:[W621 21:18:44.297861170 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.
714
+ /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.
715
+ warnings.warn(
716
+ /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.
717
+ warnings.warn(
718
+ /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.
719
+ warnings.warn(
720
+ /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.
721
+ warnings.warn(
722
+ /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.
723
+ warnings.warn(
724
+ /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.
725
+ warnings.warn(
726
+ /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.
727
+ warnings.warn(
728
+ /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.
729
+ warnings.warn(
730
+ /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.
731
+ warnings.warn(
732
+ /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.
733
+ warnings.warn(
734
+ /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.
735
+ warnings.warn(
736
+ /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.
737
+ warnings.warn(
738
+ /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.
739
+ warnings.warn(
740
+ /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.
741
+ warnings.warn(
742
+ /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.
743
+ warnings.warn(
744
+ /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.
745
+ warnings.warn(
746
+ /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.
747
+ warnings.warn(
748
+ /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.
749
+ warnings.warn(
750
+ /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.
751
+ warnings.warn(
752
+ /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.
753
+ warnings.warn(
754
+ /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.
755
+ warnings.warn(
756
+ /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.
757
+ warnings.warn(
758
+ /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.
759
+ warnings.warn(
760
+ /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.
761
+ warnings.warn(
762
+ /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.
763
+ warnings.warn(
764
+ /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.
765
+ warnings.warn(
766
+ /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.
767
+ warnings.warn(
768
+ /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.
769
+ warnings.warn(
770
+ /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.
771
+ warnings.warn(
772
+ /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.
773
+ warnings.warn(
774
+ /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.
775
+ warnings.warn(
776
+ /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.
777
+ warnings.warn(
778
+ [rank0]: Traceback (most recent call last):
779
+ [rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in <module>
780
+ [rank0]: pretrain(
781
+ [rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 879, in pretrain
782
+ [rank0]: save_checkpoint(
783
+ [rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/checkpointing.py", line 469, in save_checkpoint
784
+ [rank0]: async_save_request = dist_checkpointing.save(state_dict, checkpoint_name, save_strategy,
785
+ [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
786
+ [rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/serialization.py", line 386, in save
787
+ [rank0]: common_strategy.save_common(state_dict, checkpoint_dir)
788
+ [rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/dist_checkpointing/strategies/common.py", line 48, in save_common
789
+ [rank0]: torch.save(common_state_dict, path)
790
+ [rank0]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 964, in save
791
+ [rank0]: with _open_zipfile_writer(f) as opened_zipfile:
792
+ [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^
793
+ [rank0]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 828, in _open_zipfile_writer
794
+ [rank0]: return container(name_or_buffer)
795
+ [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^
796
+ [rank0]: File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/serialization.py", line 792, in __init__
797
+ [rank0]: torch._C.PyTorchFileWriter(
798
+ [rank0]: RuntimeError: Parent directory gpt-checkpoint/iter_0000010 does not exist.
attnserver.run_attnserver.slurm.sh.343203.out.log CHANGED
The diff for this file is too large to render. See raw diff
 
attnserver.run_attnserver.slurm.sh.343204.err.log CHANGED
The diff for this file is too large to render. See raw diff
 
attnserver.run_attnserver.slurm.sh.343204.out.log CHANGED
The diff for this file is too large to render. See raw diff
 
attnserver.run_attnserver.slurm.sh.343205.err.log CHANGED
The diff for this file is too large to render. See raw diff
 
attnserver.run_attnserver.slurm.sh.343206.err.log CHANGED
The diff for this file is too large to render. See raw diff
 
attnserver.run_attnserver.slurm.sh.343206.out.log CHANGED
The diff for this file is too large to render. See raw diff
 
attnserver.run_attnserver.slurm.sh.343207.err.log ADDED
@@ -0,0 +1,141 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ + source /mnt/weka/home/hao.zhang/conda/miniconda/bin/activate
2
+ ++ _CONDA_ROOT=/mnt/weka/home/hao.zhang/conda/miniconda
3
+ ++ . /mnt/weka/home/hao.zhang/conda/miniconda/etc/profile.d/conda.sh
4
+ +++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
5
+ +++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
6
+ +++ export _CE_M=
7
+ +++ _CE_M=
8
+ +++ export _CE_CONDA=
9
+ +++ _CE_CONDA=
10
+ +++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
11
+ +++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
12
+ +++ '[' -z x ']'
13
+ ++ conda activate
14
+ ++ local cmd=activate
15
+ ++ case "$cmd" in
16
+ ++ __conda_activate activate
17
+ ++ '[' -n '' ']'
18
+ ++ local ask_conda
19
+ +++ PS1=
20
+ +++ __conda_exe shell.posix activate
21
+ +++ '[' -n '' ']'
22
+ +++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate
23
+ ++ ask_conda='unset _CE_M
24
+ unset _CE_CONDA
25
+ PS1='\''(base) '\''
26
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
27
+ export CONDA_SHLVL='\''1'\''
28
+ export CONDA_PROMPT_MODIFIER='\''(base) '\''
29
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
30
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
31
+ ++ eval 'unset _CE_M
32
+ unset _CE_CONDA
33
+ PS1='\''(base) '\''
34
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
35
+ export CONDA_SHLVL='\''1'\''
36
+ export CONDA_PROMPT_MODIFIER='\''(base) '\''
37
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
38
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
39
+ +++ unset _CE_M
40
+ +++ unset _CE_CONDA
41
+ +++ PS1='(base) '
42
+ +++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
43
+ +++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
44
+ +++ export CONDA_SHLVL=1
45
+ +++ CONDA_SHLVL=1
46
+ +++ export 'CONDA_PROMPT_MODIFIER=(base) '
47
+ +++ CONDA_PROMPT_MODIFIER='(base) '
48
+ +++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
49
+ +++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
50
+ +++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
51
+ +++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
52
+ ++ __conda_hashr
53
+ ++ '[' -n '' ']'
54
+ ++ '[' -n '' ']'
55
+ ++ hash -r
56
+ + conda activate junda-attnserver
57
+ + local cmd=activate
58
+ + case "$cmd" in
59
+ + __conda_activate activate junda-attnserver
60
+ + '[' -n '' ']'
61
+ + local ask_conda
62
+ ++ PS1='(base) '
63
+ ++ __conda_exe shell.posix activate junda-attnserver
64
+ ++ '[' -n '' ']'
65
+ ++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate junda-attnserver
66
+ + ask_conda='unset _CE_M
67
+ unset _CE_CONDA
68
+ PS1='\''(junda-attnserver) '\''
69
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
70
+ export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\''
71
+ export CONDA_SHLVL='\''2'\''
72
+ export CONDA_DEFAULT_ENV='\''junda-attnserver'\''
73
+ export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\''
74
+ export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\''
75
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
76
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
77
+ + eval 'unset _CE_M
78
+ unset _CE_CONDA
79
+ PS1='\''(junda-attnserver) '\''
80
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
81
+ export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\''
82
+ export CONDA_SHLVL='\''2'\''
83
+ export CONDA_DEFAULT_ENV='\''junda-attnserver'\''
84
+ export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\''
85
+ export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\''
86
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
87
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
88
+ ++ unset _CE_M
89
+ ++ unset _CE_CONDA
90
+ ++ PS1='(junda-attnserver) '
91
+ ++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
92
+ ++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
93
+ ++ export CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver
94
+ ++ CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver
95
+ ++ export CONDA_SHLVL=2
96
+ ++ CONDA_SHLVL=2
97
+ ++ export CONDA_DEFAULT_ENV=junda-attnserver
98
+ ++ CONDA_DEFAULT_ENV=junda-attnserver
99
+ ++ export 'CONDA_PROMPT_MODIFIER=(junda-attnserver) '
100
+ ++ CONDA_PROMPT_MODIFIER='(junda-attnserver) '
101
+ ++ export CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda
102
+ ++ CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda
103
+ ++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
104
+ ++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
105
+ ++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
106
+ ++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
107
+ + __conda_hashr
108
+ + '[' -n '' ']'
109
+ + '[' -n '' ']'
110
+ + hash -r
111
+ + export CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
112
+ + CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
113
+ + mkdir -p /mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
114
+ + export PROF_TP_SIZE=8
115
+ + PROF_TP_SIZE=8
116
+ + export PROF_CP_SIZE=1
117
+ + PROF_CP_SIZE=1
118
+ + export PROF_BS=1
119
+ + PROF_BS=1
120
+ + for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
121
+ + export PROF_CTX_LENGTH=1024
122
+ + PROF_CTX_LENGTH=1024
123
+ + name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp8.cp1.bs1.json'
124
+ + '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp8.cp1.bs1.json' ']'
125
+ + echo 'Running ctx_length=1024, TP_SIZE=8, CP_SIZE=1, BATCH_SIZE=1'
126
+ + srun bash ./attnserver.sh
127
+ + which python3
128
+ + python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_id 343207 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-661:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 8 --context-parallel-size 1 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 1024 --max-position-embeddings 1024 --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/
129
+ /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
130
+ and will be removed in future. Use torchrun.
131
+ Note that --use-env is set by default in torchrun.
132
+ If your script expects `--local-rank` argument to be set, please
133
+ change it to read from `os.environ['LOCAL_RANK']` instead. See
134
+ https://pytorch.org/docs/stable/distributed.html#launch-utility for
135
+ further instructions
136
+
137
+ main()
138
+ W0621 21:19:26.076000 1511074 site-packages/torch/distributed/run.py:766]
139
+ W0621 21:19:26.076000 1511074 site-packages/torch/distributed/run.py:766] *****************************************
140
+ W0621 21:19:26.076000 1511074 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.
141
+ W0621 21:19:26.076000 1511074 site-packages/torch/distributed/run.py:766] *****************************************
attnserver.run_attnserver.slurm.sh.343207.out.log ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ Running ctx_length=1024, TP_SIZE=8, CP_SIZE=1, BATCH_SIZE=1
2
+ Cleaning up checkpoint directory: gpt-checkpoint
3
+ --------------------------------
4
+ CTX_LENGTH: 1024
5
+ TP_SIZE: 8
6
+ CP_SIZE: 1
7
+ CHECKPOINT_PATH: gpt-checkpoint
8
+ PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron
9
+ --------------------------------
10
+ /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
attnserver.run_attnserver.slurm.sh.343208.err.log ADDED
@@ -0,0 +1,141 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ + source /mnt/weka/home/hao.zhang/conda/miniconda/bin/activate
2
+ ++ _CONDA_ROOT=/mnt/weka/home/hao.zhang/conda/miniconda
3
+ ++ . /mnt/weka/home/hao.zhang/conda/miniconda/etc/profile.d/conda.sh
4
+ +++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
5
+ +++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
6
+ +++ export _CE_M=
7
+ +++ _CE_M=
8
+ +++ export _CE_CONDA=
9
+ +++ _CE_CONDA=
10
+ +++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
11
+ +++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
12
+ +++ '[' -z x ']'
13
+ ++ conda activate
14
+ ++ local cmd=activate
15
+ ++ case "$cmd" in
16
+ ++ __conda_activate activate
17
+ ++ '[' -n '' ']'
18
+ ++ local ask_conda
19
+ +++ PS1=
20
+ +++ __conda_exe shell.posix activate
21
+ +++ '[' -n '' ']'
22
+ +++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate
23
+ ++ ask_conda='unset _CE_M
24
+ unset _CE_CONDA
25
+ PS1='\''(base) '\''
26
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
27
+ export CONDA_SHLVL='\''1'\''
28
+ export CONDA_PROMPT_MODIFIER='\''(base) '\''
29
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
30
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
31
+ ++ eval 'unset _CE_M
32
+ unset _CE_CONDA
33
+ PS1='\''(base) '\''
34
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
35
+ export CONDA_SHLVL='\''1'\''
36
+ export CONDA_PROMPT_MODIFIER='\''(base) '\''
37
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
38
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
39
+ +++ unset _CE_M
40
+ +++ unset _CE_CONDA
41
+ +++ PS1='(base) '
42
+ +++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
43
+ +++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
44
+ +++ export CONDA_SHLVL=1
45
+ +++ CONDA_SHLVL=1
46
+ +++ export 'CONDA_PROMPT_MODIFIER=(base) '
47
+ +++ CONDA_PROMPT_MODIFIER='(base) '
48
+ +++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
49
+ +++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
50
+ +++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
51
+ +++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
52
+ ++ __conda_hashr
53
+ ++ '[' -n '' ']'
54
+ ++ '[' -n '' ']'
55
+ ++ hash -r
56
+ + conda activate junda-attnserver
57
+ + local cmd=activate
58
+ + case "$cmd" in
59
+ + __conda_activate activate junda-attnserver
60
+ + '[' -n '' ']'
61
+ + local ask_conda
62
+ ++ PS1='(base) '
63
+ ++ __conda_exe shell.posix activate junda-attnserver
64
+ ++ '[' -n '' ']'
65
+ ++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate junda-attnserver
66
+ + ask_conda='unset _CE_M
67
+ unset _CE_CONDA
68
+ PS1='\''(junda-attnserver) '\''
69
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
70
+ export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\''
71
+ export CONDA_SHLVL='\''2'\''
72
+ export CONDA_DEFAULT_ENV='\''junda-attnserver'\''
73
+ export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\''
74
+ export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\''
75
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
76
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
77
+ + eval 'unset _CE_M
78
+ unset _CE_CONDA
79
+ PS1='\''(junda-attnserver) '\''
80
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
81
+ export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\''
82
+ export CONDA_SHLVL='\''2'\''
83
+ export CONDA_DEFAULT_ENV='\''junda-attnserver'\''
84
+ export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\''
85
+ export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\''
86
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
87
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
88
+ ++ unset _CE_M
89
+ ++ unset _CE_CONDA
90
+ ++ PS1='(junda-attnserver) '
91
+ ++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
92
+ ++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
93
+ ++ export CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver
94
+ ++ CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver
95
+ ++ export CONDA_SHLVL=2
96
+ ++ CONDA_SHLVL=2
97
+ ++ export CONDA_DEFAULT_ENV=junda-attnserver
98
+ ++ CONDA_DEFAULT_ENV=junda-attnserver
99
+ ++ export 'CONDA_PROMPT_MODIFIER=(junda-attnserver) '
100
+ ++ CONDA_PROMPT_MODIFIER='(junda-attnserver) '
101
+ ++ export CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda
102
+ ++ CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda
103
+ ++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
104
+ ++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
105
+ ++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
106
+ ++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
107
+ + __conda_hashr
108
+ + '[' -n '' ']'
109
+ + '[' -n '' ']'
110
+ + hash -r
111
+ + export CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
112
+ + CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
113
+ + mkdir -p /mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
114
+ + export PROF_TP_SIZE=8
115
+ + PROF_TP_SIZE=8
116
+ + export PROF_CP_SIZE=1
117
+ + PROF_CP_SIZE=1
118
+ + export PROF_BS=2
119
+ + PROF_BS=2
120
+ + for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
121
+ + export PROF_CTX_LENGTH=1024
122
+ + PROF_CTX_LENGTH=1024
123
+ + name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp8.cp1.bs2.json'
124
+ + '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp8.cp1.bs2.json' ']'
125
+ + echo 'Running ctx_length=1024, TP_SIZE=8, CP_SIZE=1, BATCH_SIZE=2'
126
+ + srun bash ./attnserver.sh
127
+ + which python3
128
+ + python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_id 343208 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-886:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 8 --context-parallel-size 1 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 1024 --max-position-embeddings 1024 --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/
129
+ /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
130
+ and will be removed in future. Use torchrun.
131
+ Note that --use-env is set by default in torchrun.
132
+ If your script expects `--local-rank` argument to be set, please
133
+ change it to read from `os.environ['LOCAL_RANK']` instead. See
134
+ https://pytorch.org/docs/stable/distributed.html#launch-utility for
135
+ further instructions
136
+
137
+ main()
138
+ W0621 21:19:26.532000 1692678 site-packages/torch/distributed/run.py:766]
139
+ W0621 21:19:26.532000 1692678 site-packages/torch/distributed/run.py:766] *****************************************
140
+ W0621 21:19:26.532000 1692678 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.
141
+ W0621 21:19:26.532000 1692678 site-packages/torch/distributed/run.py:766] *****************************************
attnserver.run_attnserver.slurm.sh.343208.out.log ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ Running ctx_length=1024, TP_SIZE=8, CP_SIZE=1, BATCH_SIZE=2
2
+ Cleaning up checkpoint directory: gpt-checkpoint
3
+ --------------------------------
4
+ CTX_LENGTH: 1024
5
+ TP_SIZE: 8
6
+ CP_SIZE: 1
7
+ CHECKPOINT_PATH: gpt-checkpoint
8
+ PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron
9
+ --------------------------------
10
+ /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
attnserver.run_attnserver.slurm.sh.343209.err.log ADDED
@@ -0,0 +1,149 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ + source /mnt/weka/home/hao.zhang/conda/miniconda/bin/activate
2
+ ++ _CONDA_ROOT=/mnt/weka/home/hao.zhang/conda/miniconda
3
+ ++ . /mnt/weka/home/hao.zhang/conda/miniconda/etc/profile.d/conda.sh
4
+ +++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
5
+ +++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
6
+ +++ export _CE_M=
7
+ +++ _CE_M=
8
+ +++ export _CE_CONDA=
9
+ +++ _CE_CONDA=
10
+ +++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
11
+ +++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
12
+ +++ '[' -z x ']'
13
+ ++ conda activate
14
+ ++ local cmd=activate
15
+ ++ case "$cmd" in
16
+ ++ __conda_activate activate
17
+ ++ '[' -n '' ']'
18
+ ++ local ask_conda
19
+ +++ PS1=
20
+ +++ __conda_exe shell.posix activate
21
+ +++ '[' -n '' ']'
22
+ +++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate
23
+ ++ ask_conda='unset _CE_M
24
+ unset _CE_CONDA
25
+ PS1='\''(base) '\''
26
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
27
+ export CONDA_SHLVL='\''1'\''
28
+ export CONDA_PROMPT_MODIFIER='\''(base) '\''
29
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
30
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
31
+ ++ eval 'unset _CE_M
32
+ unset _CE_CONDA
33
+ PS1='\''(base) '\''
34
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
35
+ export CONDA_SHLVL='\''1'\''
36
+ export CONDA_PROMPT_MODIFIER='\''(base) '\''
37
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
38
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
39
+ +++ unset _CE_M
40
+ +++ unset _CE_CONDA
41
+ +++ PS1='(base) '
42
+ +++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
43
+ +++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
44
+ +++ export CONDA_SHLVL=1
45
+ +++ CONDA_SHLVL=1
46
+ +++ export 'CONDA_PROMPT_MODIFIER=(base) '
47
+ +++ CONDA_PROMPT_MODIFIER='(base) '
48
+ +++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
49
+ +++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
50
+ +++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
51
+ +++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
52
+ ++ __conda_hashr
53
+ ++ '[' -n '' ']'
54
+ ++ '[' -n '' ']'
55
+ ++ hash -r
56
+ + conda activate junda-attnserver
57
+ + local cmd=activate
58
+ + case "$cmd" in
59
+ + __conda_activate activate junda-attnserver
60
+ + '[' -n '' ']'
61
+ + local ask_conda
62
+ ++ PS1='(base) '
63
+ ++ __conda_exe shell.posix activate junda-attnserver
64
+ ++ '[' -n '' ']'
65
+ ++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate junda-attnserver
66
+ + ask_conda='unset _CE_M
67
+ unset _CE_CONDA
68
+ PS1='\''(junda-attnserver) '\''
69
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
70
+ export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\''
71
+ export CONDA_SHLVL='\''2'\''
72
+ export CONDA_DEFAULT_ENV='\''junda-attnserver'\''
73
+ export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\''
74
+ export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\''
75
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
76
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
77
+ + eval 'unset _CE_M
78
+ unset _CE_CONDA
79
+ PS1='\''(junda-attnserver) '\''
80
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
81
+ export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\''
82
+ export CONDA_SHLVL='\''2'\''
83
+ export CONDA_DEFAULT_ENV='\''junda-attnserver'\''
84
+ export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\''
85
+ export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\''
86
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
87
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
88
+ ++ unset _CE_M
89
+ ++ unset _CE_CONDA
90
+ ++ PS1='(junda-attnserver) '
91
+ ++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
92
+ ++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
93
+ ++ export CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver
94
+ ++ CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver
95
+ ++ export CONDA_SHLVL=2
96
+ ++ CONDA_SHLVL=2
97
+ ++ export CONDA_DEFAULT_ENV=junda-attnserver
98
+ ++ CONDA_DEFAULT_ENV=junda-attnserver
99
+ ++ export 'CONDA_PROMPT_MODIFIER=(junda-attnserver) '
100
+ ++ CONDA_PROMPT_MODIFIER='(junda-attnserver) '
101
+ ++ export CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda
102
+ ++ CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda
103
+ ++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
104
+ ++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
105
+ ++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
106
+ ++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
107
+ + __conda_hashr
108
+ + '[' -n '' ']'
109
+ + '[' -n '' ']'
110
+ + hash -r
111
+ + export CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
112
+ + CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
113
+ + mkdir -p /mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
114
+ + export PROF_TP_SIZE=8
115
+ + PROF_TP_SIZE=8
116
+ + export PROF_CP_SIZE=1
117
+ + PROF_CP_SIZE=1
118
+ + export PROF_BS=4
119
+ + PROF_BS=4
120
+ + for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
121
+ + export PROF_CTX_LENGTH=1024
122
+ + PROF_CTX_LENGTH=1024
123
+ + name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp8.cp1.bs4.json'
124
+ + '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp8.cp1.bs4.json' ']'
125
+ + echo 'Running ctx_length=1024, TP_SIZE=8, CP_SIZE=1, BATCH_SIZE=4'
126
+ + srun bash ./attnserver.sh
127
+ + which python3
128
+ + python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_id 343209 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-702:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 8 --context-parallel-size 1 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 1024 --max-position-embeddings 1024 --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/
129
+ /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
130
+ and will be removed in future. Use torchrun.
131
+ Note that --use-env is set by default in torchrun.
132
+ If your script expects `--local-rank` argument to be set, please
133
+ change it to read from `os.environ['LOCAL_RANK']` instead. See
134
+ https://pytorch.org/docs/stable/distributed.html#launch-utility for
135
+ further instructions
136
+
137
+ main()
138
+ W0621 21:19:25.037000 1978718 site-packages/torch/distributed/run.py:766]
139
+ W0621 21:19:25.037000 1978718 site-packages/torch/distributed/run.py:766] *****************************************
140
+ W0621 21:19:25.037000 1978718 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.
141
+ W0621 21:19:25.037000 1978718 site-packages/torch/distributed/run.py:766] *****************************************
142
+ [rank1]:[W621 21:19:45.103630717 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.
143
+ [rank3]:[W621 21:19:46.254546836 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.
144
+ [rank0]:[W621 21:19:46.278781300 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.
145
+ [rank4]:[W621 21:19:46.282298864 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.
146
+ [rank7]:[W621 21:19:46.283689693 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.
147
+ [rank5]:[W621 21:19:46.293795584 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.
148
+ [rank6]:[W621 21:19:46.294037250 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.
149
+ [rank2]:[W621 21:19:46.295639773 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.
attnserver.run_attnserver.slurm.sh.343209.out.log ADDED
@@ -0,0 +1,537 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Running ctx_length=1024, TP_SIZE=8, CP_SIZE=1, BATCH_SIZE=4
2
+ Cleaning up checkpoint directory: gpt-checkpoint
3
+ --------------------------------
4
+ CTX_LENGTH: 1024
5
+ TP_SIZE: 8
6
+ CP_SIZE: 1
7
+ CHECKPOINT_PATH: gpt-checkpoint
8
+ PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron
9
+ --------------------------------
10
+ /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
11
+ INFO:megatron.training.initialize:Setting logging level to 0
12
+ using world size: 8, data-parallel size: 1, context-parallel size: 1, hierarchical context-parallel sizes: Nonetensor-model-parallel size: 8, encoder-tensor-model-parallel size: 0, pipeline-model-parallel size: 1, encoder-pipeline-model-parallel size: 0
13
+ Number of virtual stages per pipeline stage: None
14
+ WARNING: Setting args.check_for_nan_in_loss_and_grad to False since dynamic loss scaling is being used
15
+ using torch.float16 for parameters ...
16
+ ------------------------ arguments ------------------------
17
+ account_for_embedding_in_pipeline_split ......... False
18
+ account_for_loss_in_pipeline_split .............. False
19
+ accumulate_allreduce_grads_in_fp32 .............. False
20
+ adam_beta1 ...................................... 0.9
21
+ adam_beta2 ...................................... 0.999
22
+ adam_eps ........................................ 1e-08
23
+ add_bias_linear ................................. True
24
+ add_position_embedding .......................... True
25
+ add_qkv_bias .................................... True
26
+ adlr_autoresume ................................. False
27
+ adlr_autoresume_interval ........................ 1000
28
+ align_grad_reduce ............................... True
29
+ align_param_gather .............................. False
30
+ app_tag_run_name ................................ None
31
+ app_tag_run_version ............................. 0.0.0
32
+ apply_layernorm_1p .............................. False
33
+ apply_query_key_layer_scaling ................... False
34
+ apply_residual_connection_post_layernorm ........ False
35
+ apply_rope_fusion ............................... False
36
+ async_save ...................................... None
37
+ async_tensor_model_parallel_allreduce ........... True
38
+ attention_backend ............................... AttnBackend.auto
39
+ attention_dropout ............................... 0.1
40
+ attention_softmax_in_fp32 ....................... False
41
+ auto_detect_ckpt_format ......................... False
42
+ barrier_with_L1_time ............................ True
43
+ bert_binary_head ................................ True
44
+ bert_embedder_type .............................. megatron
45
+ bert_load ....................................... None
46
+ bf16 ............................................ False
47
+ bias_dropout_fusion ............................. True
48
+ bias_gelu_fusion ................................ True
49
+ bias_swiglu_fusion .............................. True
50
+ biencoder_projection_dim ........................ 0
51
+ biencoder_shared_query_context_model ............ False
52
+ block_data_path ................................. None
53
+ calc_ft_timeouts ................................ False
54
+ calculate_per_token_loss ........................ False
55
+ check_for_large_grads ........................... False
56
+ check_for_nan_in_loss_and_grad .................. False
57
+ check_for_spiky_loss ............................ False
58
+ check_weight_hash_across_dp_replicas_interval ... None
59
+ ckpt_assume_constant_structure .................. False
60
+ ckpt_convert_format ............................. None
61
+ ckpt_convert_save ............................... None
62
+ ckpt_convert_update_legacy_dist_opt_format ...... False
63
+ ckpt_format ..................................... torch_dist
64
+ ckpt_fully_parallel_load ........................ False
65
+ ckpt_fully_parallel_save ........................ True
66
+ ckpt_fully_parallel_save_deprecated ............. False
67
+ ckpt_step ....................................... None
68
+ classes_fraction ................................ 1.0
69
+ clip_grad ....................................... 1.0
70
+ clone_scatter_output_in_embedding ............... True
71
+ config_logger_dir ...............................
72
+ consumed_train_samples .......................... 0
73
+ consumed_valid_samples .......................... 0
74
+ context_parallel_size ........................... 1
75
+ cp_comm_type .................................... ['p2p']
76
+ create_attention_mask_in_dataloader ............. True
77
+ cross_entropy_fusion_impl ....................... native
78
+ cross_entropy_loss_fusion ....................... False
79
+ cuda_graph_scope ................................ full
80
+ cuda_graph_warmup_steps ......................... 3
81
+ data_args_path .................................. None
82
+ data_cache_path ................................. None
83
+ data_parallel_random_init ....................... False
84
+ data_parallel_sharding_strategy ................. no_shard
85
+ data_parallel_size .............................. 1
86
+ data_path ....................................... None
87
+ data_per_class_fraction ......................... 1.0
88
+ data_sharding ................................... True
89
+ dataloader_type ................................. single
90
+ ddp_average_in_collective ....................... False
91
+ ddp_bucket_size ................................. None
92
+ ddp_num_buckets ................................. None
93
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
94
+ decoder_first_pipeline_num_layers ............... None
95
+ decoder_last_pipeline_num_layers ................ None
96
+ decoder_num_layers .............................. None
97
+ decoder_seq_length .............................. None
98
+ decoupled_lr .................................... None
99
+ decoupled_min_lr ................................ None
100
+ decrease_batch_size_if_needed ................... False
101
+ defer_embedding_wgrad_compute ................... False
102
+ deprecated_use_mcore_models ..................... False
103
+ deterministic_mode .............................. False
104
+ dino_bottleneck_size ............................ 256
105
+ dino_freeze_last_layer .......................... 1
106
+ dino_head_hidden_size ........................... 2048
107
+ dino_local_crops_number ......................... 10
108
+ dino_local_img_size ............................. 96
109
+ dino_norm_last_layer ............................ False
110
+ dino_teacher_temp ............................... 0.07
111
+ dino_warmup_teacher_temp ........................ 0.04
112
+ dino_warmup_teacher_temp_epochs ................. 30
113
+ disable_bf16_reduced_precision_matmul ........... False
114
+ disable_mamba_mem_eff_path ...................... False
115
+ disable_straggler_on_startup .................... False
116
+ dist_ckpt_format_deprecated ..................... None
117
+ dist_ckpt_strictness ............................ assume_ok_unexpected
118
+ distribute_saved_activations .................... False
119
+ distributed_backend ............................. nccl
120
+ distributed_timeout_minutes ..................... 10
121
+ embedding_path .................................. None
122
+ empty_unused_memory_level ....................... 0
123
+ enable_cuda_graph ............................... False
124
+ enable_ft_package ............................... False
125
+ enable_gloo_process_groups ...................... True
126
+ enable_msc ...................................... True
127
+ enable_one_logger ............................... True
128
+ encoder_num_layers .............................. 2
129
+ encoder_pipeline_model_parallel_size ............ 0
130
+ encoder_seq_length .............................. 1024
131
+ encoder_tensor_model_parallel_size .............. 0
132
+ end_weight_decay ................................ 0.1
133
+ eod_mask_loss ................................... False
134
+ error_injection_rate ............................ 0
135
+ error_injection_type ............................ transient_error
136
+ eval_interval ................................... 16
137
+ eval_iters ...................................... 1
138
+ evidence_data_path .............................. None
139
+ exit_duration_in_mins ........................... None
140
+ exit_interval ................................... None
141
+ exit_on_missing_checkpoint ...................... False
142
+ exit_signal_handler ............................. False
143
+ exp_avg_dtype ................................... torch.float32
144
+ exp_avg_sq_dtype ................................ torch.float32
145
+ expert_model_parallel_size ...................... 1
146
+ expert_tensor_parallel_size ..................... 8
147
+ external_cuda_graph ............................. False
148
+ ffn_hidden_size ................................. 16384
149
+ finetune ........................................ False
150
+ first_last_layers_bf16 .......................... False
151
+ flash_decode .................................... False
152
+ fp16 ............................................ True
153
+ fp16_lm_cross_entropy ........................... False
154
+ fp32_residual_connection ........................ False
155
+ fp8 ............................................. None
156
+ fp8_amax_compute_algo ........................... most_recent
157
+ fp8_amax_history_len ............................ 1
158
+ fp8_interval .................................... 1
159
+ fp8_margin ...................................... 0
160
+ fp8_param_gather ................................ False
161
+ fp8_recipe ...................................... delayed
162
+ fp8_wgrad ....................................... True
163
+ fsdp_double_buffer .............................. False
164
+ global_batch_size ............................... 1
165
+ grad_reduce_in_bf16 ............................. False
166
+ gradient_accumulation_fusion .................... True
167
+ gradient_reduce_div_fusion ...................... True
168
+ group_query_attention ........................... True
169
+ head_lr_mult .................................... 1.0
170
+ heterogeneous_layers_config_encoded_json ........ None
171
+ heterogeneous_layers_config_path ................ None
172
+ hidden_dropout .................................. 0.1
173
+ hidden_size ..................................... 4096
174
+ hierarchical_context_parallel_sizes ............. None
175
+ high_priority_stream_groups ..................... []
176
+ hybrid_attention_ratio .......................... 0.0
177
+ hybrid_mlp_ratio ................................ 0.0
178
+ hybrid_override_pattern ......................... None
179
+ hysteresis ...................................... 2
180
+ ict_head_size ................................... None
181
+ ict_load ........................................ None
182
+ img_h ........................................... 224
183
+ img_w ........................................... 224
184
+ indexer_batch_size .............................. 128
185
+ indexer_log_interval ............................ 1000
186
+ inference_batch_times_seqlen_threshold .......... -1
187
+ inference_dynamic_batching ...................... False
188
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
189
+ inference_dynamic_batching_buffer_overflow_factor None
190
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
191
+ inference_dynamic_batching_chunk_size ........... 256
192
+ inference_dynamic_batching_max_requests_override None
193
+ inference_dynamic_batching_max_tokens_override .. None
194
+ inference_max_batch_size ........................ 8
195
+ inference_max_seq_length ........................ 2560
196
+ inference_rng_tracker ........................... False
197
+ init_method_std ................................. 0.02
198
+ init_method_xavier_uniform ...................... False
199
+ init_model_with_meta_device ..................... False
200
+ initial_loss_scale .............................. 4294967296
201
+ inprocess_active_world_size ..................... 8
202
+ inprocess_barrier_timeout ....................... 120
203
+ inprocess_completion_timeout .................... 120
204
+ inprocess_empty_cuda_cache ...................... False
205
+ inprocess_granularity ........................... node
206
+ inprocess_hard_timeout .......................... 90
207
+ inprocess_heartbeat_interval .................... 30
208
+ inprocess_heartbeat_timeout ..................... 60
209
+ inprocess_last_call_wait ........................ 1
210
+ inprocess_max_iterations ........................ None
211
+ inprocess_monitor_process_interval .............. 1.0
212
+ inprocess_monitor_thread_interval ............... 1.0
213
+ inprocess_progress_watchdog_interval ............ 1.0
214
+ inprocess_restart ............................... False
215
+ inprocess_soft_timeout .......................... 60
216
+ inprocess_termination_grace_time ................ 1
217
+ is_hybrid_model ................................. False
218
+ iter_per_epoch .................................. 1250
219
+ iterations_to_skip .............................. []
220
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
221
+ kv_channels ..................................... 64
222
+ kv_lora_rank .................................... 32
223
+ lazy_mpu_init ................................... None
224
+ load ............................................ gpt-checkpoint
225
+ load_model_opt_format ........................... False
226
+ local_rank ...................................... 0
227
+ log_interval .................................... 1
228
+ log_loss_scale_to_tensorboard ................... True
229
+ log_memory_to_tensorboard ....................... False
230
+ log_num_zeros_in_grad ........................... False
231
+ log_params_norm ................................. False
232
+ log_progress .................................... False
233
+ log_straggler ................................... False
234
+ log_throughput .................................. False
235
+ log_timers_to_tensorboard ....................... False
236
+ log_validation_ppl_to_tensorboard ............... False
237
+ log_world_size_to_tensorboard ................... False
238
+ logging_level ................................... 0
239
+ loss_scale ...................................... None
240
+ loss_scale_window ............................... 1000
241
+ lr .............................................. 0.0005
242
+ lr_decay_iters .................................. 150000
243
+ lr_decay_samples ................................ None
244
+ lr_decay_style .................................. cosine
245
+ lr_warmup_fraction .............................. None
246
+ lr_warmup_init .................................. 0.0
247
+ lr_warmup_iters ................................. 2
248
+ lr_warmup_samples ............................... 0
249
+ lr_wsd_decay_iters .............................. None
250
+ lr_wsd_decay_samples ............................ None
251
+ lr_wsd_decay_style .............................. exponential
252
+ main_grads_dtype ................................ torch.float32
253
+ main_params_dtype ............................... torch.float32
254
+ make_vocab_size_divisible_by .................... 128
255
+ mamba_head_dim .................................. 64
256
+ mamba_num_groups ................................ 8
257
+ mamba_num_heads ................................. None
258
+ mamba_state_dim ................................. 128
259
+ manual_gc ....................................... False
260
+ manual_gc_eval .................................. True
261
+ manual_gc_interval .............................. 0
262
+ mask_factor ..................................... 1.0
263
+ mask_prob ....................................... 0.15
264
+ mask_type ....................................... random
265
+ masked_softmax_fusion ........................... True
266
+ max_position_embeddings ......................... 1024
267
+ max_tokens_to_oom ............................... 12000
268
+ memory_snapshot_path ............................ snapshot.pickle
269
+ merge_file ...................................... merges.txt
270
+ micro_batch_size ................................ 1
271
+ microbatch_group_size_per_vp_stage .............. None
272
+ mid_level_dataset_surplus ....................... 0.005
273
+ min_loss_scale .................................. 1.0
274
+ min_lr .......................................... 0.0
275
+ mlp_chunks_for_prefill .......................... 1
276
+ mmap_bin_files .................................. True
277
+ mock_data ....................................... True
278
+ moe_apply_probs_on_input ........................ False
279
+ moe_aux_loss_coeff .............................. 0.0
280
+ moe_enable_deepep ............................... False
281
+ moe_expert_capacity_factor ...................... None
282
+ moe_extended_tp ................................. False
283
+ moe_ffn_hidden_size ............................. None
284
+ moe_grouped_gemm ................................ False
285
+ moe_input_jitter_eps ............................ None
286
+ moe_layer_freq .................................. 1
287
+ moe_layer_recompute ............................. False
288
+ moe_pad_expert_input_to_capacity ................ False
289
+ moe_per_layer_logging ........................... False
290
+ moe_permute_fusion .............................. False
291
+ moe_router_bias_update_rate ..................... 0.001
292
+ moe_router_dtype ................................ None
293
+ moe_router_enable_expert_bias ................... False
294
+ moe_router_force_load_balancing ................. False
295
+ moe_router_group_topk ........................... None
296
+ moe_router_load_balancing_type .................. aux_loss
297
+ moe_router_num_groups ........................... None
298
+ moe_router_padding_for_fp8 ...................... False
299
+ moe_router_pre_softmax .......................... False
300
+ moe_router_score_function ....................... softmax
301
+ moe_router_topk ................................. 2
302
+ moe_router_topk_scaling_factor .................. None
303
+ moe_shared_expert_intermediate_size ............. None
304
+ moe_shared_expert_overlap ....................... False
305
+ moe_token_dispatcher_type ....................... allgather
306
+ moe_token_drop_policy ........................... probs
307
+ moe_use_legacy_grouped_gemm ..................... False
308
+ moe_use_upcycling ............................... False
309
+ moe_z_loss_coeff ................................ None
310
+ mrope_section ................................... None
311
+ mscale .......................................... 1.0
312
+ mscale_all_dim .................................. 1.0
313
+ mtp_loss_scaling_factor ......................... 0.1
314
+ mtp_num_layers .................................. None
315
+ multi_latent_attention .......................... False
316
+ nccl_all_reduce_for_prefill ..................... False
317
+ nccl_communicator_config_path ................... None
318
+ nccl_ub ......................................... False
319
+ no_load_optim ................................... None
320
+ no_load_rng ..................................... None
321
+ no_persist_layer_norm ........................... False
322
+ no_rope_freq .................................... None
323
+ no_save_optim ................................... None
324
+ no_save_rng ..................................... None
325
+ non_persistent_ckpt_type ........................ None
326
+ non_persistent_global_ckpt_dir .................. None
327
+ non_persistent_local_ckpt_algo .................. fully_parallel
328
+ non_persistent_local_ckpt_dir ................... None
329
+ non_persistent_save_interval .................... None
330
+ norm_epsilon .................................... 1e-05
331
+ normalization ................................... LayerNorm
332
+ num_attention_heads ............................. 64
333
+ num_channels .................................... 3
334
+ num_classes ..................................... 1000
335
+ num_dataset_builder_threads ..................... 1
336
+ num_distributed_optimizer_instances ............. 1
337
+ num_experts ..................................... None
338
+ num_layers ...................................... 2
339
+ num_layers_at_end_in_bf16 ....................... 1
340
+ num_layers_at_start_in_bf16 ..................... 1
341
+ num_layers_per_virtual_pipeline_stage ........... None
342
+ num_query_groups ................................ 16
343
+ num_virtual_stages_per_pipeline_rank ............ None
344
+ num_workers ..................................... 2
345
+ object_storage_cache_path ....................... None
346
+ one_logger_async ................................ False
347
+ one_logger_project .............................. megatron-lm
348
+ one_logger_run_name ............................. None
349
+ onnx_safe ....................................... None
350
+ openai_gelu ..................................... False
351
+ optimizer ....................................... adam
352
+ optimizer_cpu_offload ........................... False
353
+ optimizer_offload_fraction ...................... 1.0
354
+ output_bert_embeddings .......................... False
355
+ overlap_cpu_optimizer_d2h_h2d ................... False
356
+ overlap_grad_reduce ............................. False
357
+ overlap_p2p_comm ................................ False
358
+ overlap_p2p_comm_warmup_flush ................... False
359
+ overlap_param_gather ............................ False
360
+ overlap_param_gather_with_optimizer_step ........ False
361
+ override_opt_param_scheduler .................... False
362
+ params_dtype .................................... torch.float16
363
+ patch_dim ....................................... 16
364
+ per_split_data_args_path ........................ None
365
+ perform_initialization .......................... True
366
+ pin_cpu_grads ................................... True
367
+ pin_cpu_params .................................. True
368
+ pipeline_model_parallel_comm_backend ............ None
369
+ pipeline_model_parallel_size .................... 1
370
+ pipeline_model_parallel_split_rank .............. None
371
+ position_embedding_type ......................... learned_absolute
372
+ pretrained_checkpoint ........................... None
373
+ profile ......................................... False
374
+ profile_ranks ................................... [0]
375
+ profile_step_end ................................ 12
376
+ profile_step_start .............................. 10
377
+ q_lora_rank ..................................... None
378
+ qk_head_dim ..................................... 128
379
+ qk_l2_norm ...................................... False
380
+ qk_layernorm .................................... False
381
+ qk_pos_emb_head_dim ............................. 64
382
+ query_in_block_prob ............................. 0.1
383
+ rampup_batch_size ............................... None
384
+ rank ............................................ 0
385
+ recompute_granularity ........................... None
386
+ recompute_method ................................ None
387
+ recompute_modules ............................... None
388
+ recompute_num_layers ............................ None
389
+ record_memory_history ........................... False
390
+ relative_attention_max_distance ................. 128
391
+ relative_attention_num_buckets .................. 32
392
+ replication ..................................... False
393
+ replication_factor .............................. 2
394
+ replication_jump ................................ None
395
+ rerun_mode ...................................... disabled
396
+ reset_attention_mask ............................ False
397
+ reset_position_ids .............................. False
398
+ result_rejected_tracker_filename ................ None
399
+ retriever_report_topk_accuracies ................ []
400
+ retriever_score_scaling ......................... False
401
+ retriever_seq_length ............................ 256
402
+ retro_add_retriever ............................. False
403
+ retro_attention_gate ............................ 1
404
+ retro_cyclic_train_iters ........................ None
405
+ retro_encoder_attention_dropout ................. 0.1
406
+ retro_encoder_hidden_dropout .................... 0.1
407
+ retro_encoder_layers ............................ 2
408
+ retro_num_neighbors ............................. 2
409
+ retro_num_retrieved_chunks ...................... 2
410
+ retro_project_dir ............................... None
411
+ retro_verify_neighbor_count ..................... True
412
+ rope_scaling_factor ............................. 8.0
413
+ rotary_base ..................................... 10000
414
+ rotary_interleaved .............................. False
415
+ rotary_percent .................................. 1.0
416
+ rotary_scaling_factor ........................... 1.0
417
+ rotary_seq_len_interpolation_factor ............. None
418
+ run_workload_inspector_server ................... False
419
+ sample_rate ..................................... 1.0
420
+ save ............................................ gpt-checkpoint
421
+ save_interval ................................... 16
422
+ scatter_gather_tensors_in_pipeline .............. True
423
+ seed ............................................ 1234
424
+ seq_length ...................................... 1024
425
+ sequence_parallel ............................... False
426
+ sgd_momentum .................................... 0.9
427
+ short_seq_prob .................................. 0.1
428
+ skip_train ...................................... False
429
+ skipped_train_samples ........................... 0
430
+ spec ............................................ None
431
+ split ........................................... None
432
+ squared_relu .................................... False
433
+ start_weight_decay .............................. 0.1
434
+ straggler_ctrlr_port ............................ 65535
435
+ straggler_minmax_count .......................... 1
436
+ suggested_communication_unit_size ............... None
437
+ swiglu .......................................... False
438
+ swin_backbone_type .............................. tiny
439
+ symmetric_ar_type ............................... None
440
+ te_rng_tracker .................................. False
441
+ tensor_model_parallel_size ...................... 8
442
+ tensorboard_dir ................................. tensorboard-logs/
443
+ tensorboard_log_interval ........................ 1
444
+ tensorboard_queue_size .......................... 1000
445
+ test_data_path .................................. None
446
+ test_mode ....................................... False
447
+ tiktoken_num_special_tokens ..................... 1000
448
+ tiktoken_pattern ................................ None
449
+ tiktoken_special_tokens ......................... None
450
+ timing_log_level ................................ 0
451
+ timing_log_option ............................... minmax
452
+ titles_data_path ................................ None
453
+ tokenizer_model ................................. None
454
+ tokenizer_type .................................. GPT2BPETokenizer
455
+ torch_fsdp2_reshard_after_forward ............... True
456
+ tp_comm_bootstrap_backend ....................... nccl
457
+ tp_comm_bulk_dgrad .............................. True
458
+ tp_comm_bulk_wgrad .............................. True
459
+ tp_comm_overlap ................................. False
460
+ tp_comm_overlap_ag .............................. True
461
+ tp_comm_overlap_cfg ............................. None
462
+ tp_comm_overlap_rs .............................. True
463
+ tp_comm_overlap_rs_dgrad ........................ False
464
+ tp_comm_split_ag ................................ True
465
+ tp_comm_split_rs ................................ True
466
+ train_data_path ................................. None
467
+ train_iters ..................................... 10
468
+ train_samples ................................... None
469
+ train_sync_interval ............................. None
470
+ transformer_impl ................................ transformer_engine
471
+ transformer_pipeline_model_parallel_size ........ 1
472
+ untie_embeddings_and_output_weights ............. False
473
+ use_checkpoint_args ............................. False
474
+ use_checkpoint_opt_param_scheduler .............. False
475
+ use_cpu_initialization .......................... None
476
+ use_custom_fsdp ................................. False
477
+ use_dist_ckpt ................................... True
478
+ use_dist_ckpt_deprecated ........................ False
479
+ use_distributed_optimizer ....................... False
480
+ use_flash_attn .................................. False
481
+ use_legacy_models ............................... False
482
+ use_mp_args_from_checkpoint_args ................ False
483
+ use_one_sent_docs ............................... False
484
+ use_persistent_ckpt_worker ...................... False
485
+ use_precision_aware_optimizer ................... False
486
+ use_pytorch_profiler ............................ False
487
+ use_ring_exchange_p2p ........................... False
488
+ use_rope_scaling ................................ False
489
+ use_rotary_position_embeddings .................. False
490
+ use_sharp ....................................... False
491
+ use_tokenizer_model_from_checkpoint_args ........ True
492
+ use_torch_fsdp2 ................................. False
493
+ use_torch_optimizer_for_cpu_offload ............. False
494
+ use_tp_pp_dp_mapping ............................ False
495
+ v_head_dim ...................................... 128
496
+ valid_data_path ................................. None
497
+ variable_seq_lengths ............................ False
498
+ virtual_pipeline_model_parallel_size ............ None
499
+ vision_backbone_type ............................ vit
500
+ vision_pretraining .............................. False
501
+ vision_pretraining_type ......................... classify
502
+ vocab_extra_ids ................................. 0
503
+ vocab_file ...................................... vocab.json
504
+ vocab_size ...................................... None
505
+ wandb_exp_name ..................................
506
+ wandb_project ...................................
507
+ wandb_save_dir ..................................
508
+ weight_decay .................................... 0.1
509
+ weight_decay_incr_style ......................... constant
510
+ wgrad_deferral_limit ............................ 0
511
+ world_size ...................................... 8
512
+ yaml_cfg ........................................ None
513
+ -------------------- end of arguments ---------------------
514
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
515
+ > building GPT2BPETokenizer tokenizer ...
516
+ INFO:megatron.training.initialize:Setting logging level to 0
517
+ > padded vocab (size: 50257) with 943 dummy tokens (new size: 51200)
518
+ INFO:megatron.training.initialize:Setting logging level to 0
519
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.DISABLED
520
+ > initializing torch distributed ...
521
+ > initialized tensor model parallel with size 8
522
+ > initialized pipeline model parallel with size 1
523
+ > setting random seeds to 1234 ...
524
+ > compiling dataset index builder ...
525
+ make: Entering directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets'
526
+ INFO:megatron.training.initialize:Setting logging level to 0
527
+ WARNING: TensorBoard writing requested but is not available (are you using PyTorch 1.1.0 or later?), no TensorBoard logs will be written.
528
+ WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it
529
+ INFO:megatron.training.initialize:Setting logging level to 0
530
+ INFO:megatron.training.initialize:Setting logging level to 0
531
+ INFO:megatron.training.initialize:Setting logging level to 0
532
+ make: Nothing to be done for 'default'.
533
+ make: Leaving directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets'
534
+ >>> done with dataset index builder. Compilation time: 0.044 seconds
535
+ > compiling and loading fused kernels ...
536
+ INFO:megatron.training.initialize:Setting logging level to 0
537
+ >>> done with compiling and loading fused kernels. Compilation time: 2.457 seconds
attnserver.run_attnserver.slurm.sh.343210.err.log ADDED
@@ -0,0 +1,149 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ + source /mnt/weka/home/hao.zhang/conda/miniconda/bin/activate
2
+ ++ _CONDA_ROOT=/mnt/weka/home/hao.zhang/conda/miniconda
3
+ ++ . /mnt/weka/home/hao.zhang/conda/miniconda/etc/profile.d/conda.sh
4
+ +++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
5
+ +++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
6
+ +++ export _CE_M=
7
+ +++ _CE_M=
8
+ +++ export _CE_CONDA=
9
+ +++ _CE_CONDA=
10
+ +++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
11
+ +++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
12
+ +++ '[' -z x ']'
13
+ ++ conda activate
14
+ ++ local cmd=activate
15
+ ++ case "$cmd" in
16
+ ++ __conda_activate activate
17
+ ++ '[' -n '' ']'
18
+ ++ local ask_conda
19
+ +++ PS1=
20
+ +++ __conda_exe shell.posix activate
21
+ +++ '[' -n '' ']'
22
+ +++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate
23
+ ++ ask_conda='unset _CE_M
24
+ unset _CE_CONDA
25
+ PS1='\''(base) '\''
26
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
27
+ export CONDA_SHLVL='\''1'\''
28
+ export CONDA_PROMPT_MODIFIER='\''(base) '\''
29
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
30
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
31
+ ++ eval 'unset _CE_M
32
+ unset _CE_CONDA
33
+ PS1='\''(base) '\''
34
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
35
+ export CONDA_SHLVL='\''1'\''
36
+ export CONDA_PROMPT_MODIFIER='\''(base) '\''
37
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
38
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
39
+ +++ unset _CE_M
40
+ +++ unset _CE_CONDA
41
+ +++ PS1='(base) '
42
+ +++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
43
+ +++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
44
+ +++ export CONDA_SHLVL=1
45
+ +++ CONDA_SHLVL=1
46
+ +++ export 'CONDA_PROMPT_MODIFIER=(base) '
47
+ +++ CONDA_PROMPT_MODIFIER='(base) '
48
+ +++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
49
+ +++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
50
+ +++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
51
+ +++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
52
+ ++ __conda_hashr
53
+ ++ '[' -n '' ']'
54
+ ++ '[' -n '' ']'
55
+ ++ hash -r
56
+ + conda activate junda-attnserver
57
+ + local cmd=activate
58
+ + case "$cmd" in
59
+ + __conda_activate activate junda-attnserver
60
+ + '[' -n '' ']'
61
+ + local ask_conda
62
+ ++ PS1='(base) '
63
+ ++ __conda_exe shell.posix activate junda-attnserver
64
+ ++ '[' -n '' ']'
65
+ ++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate junda-attnserver
66
+ + ask_conda='unset _CE_M
67
+ unset _CE_CONDA
68
+ PS1='\''(junda-attnserver) '\''
69
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
70
+ export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\''
71
+ export CONDA_SHLVL='\''2'\''
72
+ export CONDA_DEFAULT_ENV='\''junda-attnserver'\''
73
+ export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\''
74
+ export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\''
75
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
76
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
77
+ + eval 'unset _CE_M
78
+ unset _CE_CONDA
79
+ PS1='\''(junda-attnserver) '\''
80
+ export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\''
81
+ export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\''
82
+ export CONDA_SHLVL='\''2'\''
83
+ export CONDA_DEFAULT_ENV='\''junda-attnserver'\''
84
+ export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\''
85
+ export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\''
86
+ export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\''
87
+ export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\'''
88
+ ++ unset _CE_M
89
+ ++ unset _CE_CONDA
90
+ ++ PS1='(junda-attnserver) '
91
+ ++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
92
+ ++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
93
+ ++ export CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver
94
+ ++ CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver
95
+ ++ export CONDA_SHLVL=2
96
+ ++ CONDA_SHLVL=2
97
+ ++ export CONDA_DEFAULT_ENV=junda-attnserver
98
+ ++ CONDA_DEFAULT_ENV=junda-attnserver
99
+ ++ export 'CONDA_PROMPT_MODIFIER=(junda-attnserver) '
100
+ ++ CONDA_PROMPT_MODIFIER='(junda-attnserver) '
101
+ ++ export CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda
102
+ ++ CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda
103
+ ++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
104
+ ++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda
105
+ ++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
106
+ ++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python
107
+ + __conda_hashr
108
+ + '[' -n '' ']'
109
+ + '[' -n '' ']'
110
+ + hash -r
111
+ + export CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
112
+ + CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
113
+ + mkdir -p /mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5
114
+ + export PROF_TP_SIZE=8
115
+ + PROF_TP_SIZE=8
116
+ + export PROF_CP_SIZE=1
117
+ + PROF_CP_SIZE=1
118
+ + export PROF_BS=8
119
+ + PROF_BS=8
120
+ + for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072
121
+ + export PROF_CTX_LENGTH=1024
122
+ + PROF_CTX_LENGTH=1024
123
+ + name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp8.cp1.bs8.json'
124
+ + '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp8.cp1.bs8.json' ']'
125
+ + echo 'Running ctx_length=1024, TP_SIZE=8, CP_SIZE=1, BATCH_SIZE=8'
126
+ + srun bash ./attnserver.sh
127
+ + which python3
128
+ + python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_id 343210 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-768:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 8 --context-parallel-size 1 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 1024 --max-position-embeddings 1024 --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/
129
+ /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
130
+ and will be removed in future. Use torchrun.
131
+ Note that --use-env is set by default in torchrun.
132
+ If your script expects `--local-rank` argument to be set, please
133
+ change it to read from `os.environ['LOCAL_RANK']` instead. See
134
+ https://pytorch.org/docs/stable/distributed.html#launch-utility for
135
+ further instructions
136
+
137
+ main()
138
+ W0621 21:19:24.688000 2174748 site-packages/torch/distributed/run.py:766]
139
+ W0621 21:19:24.688000 2174748 site-packages/torch/distributed/run.py:766] *****************************************
140
+ W0621 21:19:24.688000 2174748 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.
141
+ W0621 21:19:24.688000 2174748 site-packages/torch/distributed/run.py:766] *****************************************
142
+ [rank1]:[W621 21:19:46.843638124 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.
143
+ [rank3]:[W621 21:19:46.273254932 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.
144
+ [rank7]:[W621 21:19:46.287551098 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.
145
+ [rank0]:[W621 21:19:46.300492516 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.
146
+ [rank5]:[W621 21:19:46.306446014 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.
147
+ [rank4]:[W621 21:19:46.306935911 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.
148
+ [rank6]:[W621 21:19:46.310121545 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.
149
+ [rank2]:[W621 21:19:46.317946415 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.
attnserver.run_attnserver.slurm.sh.343210.out.log ADDED
@@ -0,0 +1,536 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Running ctx_length=1024, TP_SIZE=8, CP_SIZE=1, BATCH_SIZE=8
2
+ Cleaning up checkpoint directory: gpt-checkpoint
3
+ --------------------------------
4
+ CTX_LENGTH: 1024
5
+ TP_SIZE: 8
6
+ CP_SIZE: 1
7
+ CHECKPOINT_PATH: gpt-checkpoint
8
+ PWD: /mnt/weka/home/hao.zhang/junda/attnserver-megatron
9
+ --------------------------------
10
+ /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3
11
+ INFO:megatron.training.initialize:Setting logging level to 0
12
+ using world size: 8, data-parallel size: 1, context-parallel size: 1, hierarchical context-parallel sizes: Nonetensor-model-parallel size: 8, encoder-tensor-model-parallel size: 0, pipeline-model-parallel size: 1, encoder-pipeline-model-parallel size: 0
13
+ Number of virtual stages per pipeline stage: None
14
+ WARNING: Setting args.check_for_nan_in_loss_and_grad to False since dynamic loss scaling is being used
15
+ using torch.float16 for parameters ...
16
+ ------------------------ arguments ------------------------
17
+ account_for_embedding_in_pipeline_split ......... False
18
+ account_for_loss_in_pipeline_split .............. False
19
+ accumulate_allreduce_grads_in_fp32 .............. False
20
+ adam_beta1 ...................................... 0.9
21
+ adam_beta2 ...................................... 0.999
22
+ adam_eps ........................................ 1e-08
23
+ add_bias_linear ................................. True
24
+ add_position_embedding .......................... True
25
+ add_qkv_bias .................................... True
26
+ adlr_autoresume ................................. False
27
+ adlr_autoresume_interval ........................ 1000
28
+ align_grad_reduce ............................... True
29
+ align_param_gather .............................. False
30
+ app_tag_run_name ................................ None
31
+ app_tag_run_version ............................. 0.0.0
32
+ apply_layernorm_1p .............................. False
33
+ apply_query_key_layer_scaling ................... False
34
+ apply_residual_connection_post_layernorm ........ False
35
+ apply_rope_fusion ............................... False
36
+ async_save ...................................... None
37
+ async_tensor_model_parallel_allreduce ........... True
38
+ attention_backend ............................... AttnBackend.auto
39
+ attention_dropout ............................... 0.1
40
+ attention_softmax_in_fp32 ....................... False
41
+ auto_detect_ckpt_format ......................... False
42
+ barrier_with_L1_time ............................ True
43
+ bert_binary_head ................................ True
44
+ bert_embedder_type .............................. megatron
45
+ bert_load ....................................... None
46
+ bf16 ............................................ False
47
+ bias_dropout_fusion ............................. True
48
+ bias_gelu_fusion ................................ True
49
+ bias_swiglu_fusion .............................. True
50
+ biencoder_projection_dim ........................ 0
51
+ biencoder_shared_query_context_model ............ False
52
+ block_data_path ................................. None
53
+ calc_ft_timeouts ................................ False
54
+ calculate_per_token_loss ........................ False
55
+ check_for_large_grads ........................... False
56
+ check_for_nan_in_loss_and_grad .................. False
57
+ check_for_spiky_loss ............................ False
58
+ check_weight_hash_across_dp_replicas_interval ... None
59
+ ckpt_assume_constant_structure .................. False
60
+ ckpt_convert_format ............................. None
61
+ ckpt_convert_save ............................... None
62
+ ckpt_convert_update_legacy_dist_opt_format ...... False
63
+ ckpt_format ..................................... torch_dist
64
+ ckpt_fully_parallel_load ........................ False
65
+ ckpt_fully_parallel_save ........................ True
66
+ ckpt_fully_parallel_save_deprecated ............. False
67
+ ckpt_step ....................................... None
68
+ classes_fraction ................................ 1.0
69
+ clip_grad ....................................... 1.0
70
+ clone_scatter_output_in_embedding ............... True
71
+ config_logger_dir ...............................
72
+ consumed_train_samples .......................... 0
73
+ consumed_valid_samples .......................... 0
74
+ context_parallel_size ........................... 1
75
+ cp_comm_type .................................... ['p2p']
76
+ create_attention_mask_in_dataloader ............. True
77
+ cross_entropy_fusion_impl ....................... native
78
+ cross_entropy_loss_fusion ....................... False
79
+ cuda_graph_scope ................................ full
80
+ cuda_graph_warmup_steps ......................... 3
81
+ data_args_path .................................. None
82
+ data_cache_path ................................. None
83
+ data_parallel_random_init ....................... False
84
+ data_parallel_sharding_strategy ................. no_shard
85
+ data_parallel_size .............................. 1
86
+ data_path ....................................... None
87
+ data_per_class_fraction ......................... 1.0
88
+ data_sharding ................................... True
89
+ dataloader_type ................................. single
90
+ ddp_average_in_collective ....................... False
91
+ ddp_bucket_size ................................. None
92
+ ddp_num_buckets ................................. None
93
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
94
+ decoder_first_pipeline_num_layers ............... None
95
+ decoder_last_pipeline_num_layers ................ None
96
+ decoder_num_layers .............................. None
97
+ decoder_seq_length .............................. None
98
+ decoupled_lr .................................... None
99
+ decoupled_min_lr ................................ None
100
+ decrease_batch_size_if_needed ................... False
101
+ defer_embedding_wgrad_compute ................... False
102
+ deprecated_use_mcore_models ..................... False
103
+ deterministic_mode .............................. False
104
+ dino_bottleneck_size ............................ 256
105
+ dino_freeze_last_layer .......................... 1
106
+ dino_head_hidden_size ........................... 2048
107
+ dino_local_crops_number ......................... 10
108
+ dino_local_img_size ............................. 96
109
+ dino_norm_last_layer ............................ False
110
+ dino_teacher_temp ............................... 0.07
111
+ dino_warmup_teacher_temp ........................ 0.04
112
+ dino_warmup_teacher_temp_epochs ................. 30
113
+ disable_bf16_reduced_precision_matmul ........... False
114
+ disable_mamba_mem_eff_path ...................... False
115
+ disable_straggler_on_startup .................... False
116
+ dist_ckpt_format_deprecated ..................... None
117
+ dist_ckpt_strictness ............................ assume_ok_unexpected
118
+ distribute_saved_activations .................... False
119
+ distributed_backend ............................. nccl
120
+ distributed_timeout_minutes ..................... 10
121
+ embedding_path .................................. None
122
+ empty_unused_memory_level ....................... 0
123
+ enable_cuda_graph ............................... False
124
+ enable_ft_package ............................... False
125
+ enable_gloo_process_groups ...................... True
126
+ enable_msc ...................................... True
127
+ enable_one_logger ............................... True
128
+ encoder_num_layers .............................. 2
129
+ encoder_pipeline_model_parallel_size ............ 0
130
+ encoder_seq_length .............................. 1024
131
+ encoder_tensor_model_parallel_size .............. 0
132
+ end_weight_decay ................................ 0.1
133
+ eod_mask_loss ................................... False
134
+ error_injection_rate ............................ 0
135
+ error_injection_type ............................ transient_error
136
+ eval_interval ................................... 16
137
+ eval_iters ...................................... 1
138
+ evidence_data_path .............................. None
139
+ exit_duration_in_mins ........................... None
140
+ exit_interval ................................... None
141
+ exit_on_missing_checkpoint ...................... False
142
+ exit_signal_handler ............................. False
143
+ exp_avg_dtype ................................... torch.float32
144
+ exp_avg_sq_dtype ................................ torch.float32
145
+ expert_model_parallel_size ...................... 1
146
+ expert_tensor_parallel_size ..................... 8
147
+ external_cuda_graph ............................. False
148
+ ffn_hidden_size ................................. 16384
149
+ finetune ........................................ False
150
+ first_last_layers_bf16 .......................... False
151
+ flash_decode .................................... False
152
+ fp16 ............................................ True
153
+ fp16_lm_cross_entropy ........................... False
154
+ fp32_residual_connection ........................ False
155
+ fp8 ............................................. None
156
+ fp8_amax_compute_algo ........................... most_recent
157
+ fp8_amax_history_len ............................ 1
158
+ fp8_interval .................................... 1
159
+ fp8_margin ...................................... 0
160
+ fp8_param_gather ................................ False
161
+ fp8_recipe ...................................... delayed
162
+ fp8_wgrad ....................................... True
163
+ fsdp_double_buffer .............................. False
164
+ global_batch_size ............................... 1
165
+ grad_reduce_in_bf16 ............................. False
166
+ gradient_accumulation_fusion .................... True
167
+ gradient_reduce_div_fusion ...................... True
168
+ group_query_attention ........................... True
169
+ head_lr_mult .................................... 1.0
170
+ heterogeneous_layers_config_encoded_json ........ None
171
+ heterogeneous_layers_config_path ................ None
172
+ hidden_dropout .................................. 0.1
173
+ hidden_size ..................................... 4096
174
+ hierarchical_context_parallel_sizes ............. None
175
+ high_priority_stream_groups ..................... []
176
+ hybrid_attention_ratio .......................... 0.0
177
+ hybrid_mlp_ratio ................................ 0.0
178
+ hybrid_override_pattern ......................... None
179
+ hysteresis ...................................... 2
180
+ ict_head_size ................................... None
181
+ ict_load ........................................ None
182
+ img_h ........................................... 224
183
+ img_w ........................................... 224
184
+ indexer_batch_size .............................. 128
185
+ indexer_log_interval ............................ 1000
186
+ inference_batch_times_seqlen_threshold .......... -1
187
+ inference_dynamic_batching ...................... False
188
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
189
+ inference_dynamic_batching_buffer_overflow_factor None
190
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
191
+ inference_dynamic_batching_chunk_size ........... 256
192
+ inference_dynamic_batching_max_requests_override None
193
+ inference_dynamic_batching_max_tokens_override .. None
194
+ inference_max_batch_size ........................ 8
195
+ inference_max_seq_length ........................ 2560
196
+ inference_rng_tracker ........................... False
197
+ init_method_std ................................. 0.02
198
+ init_method_xavier_uniform ...................... False
199
+ init_model_with_meta_device ..................... False
200
+ initial_loss_scale .............................. 4294967296
201
+ inprocess_active_world_size ..................... 8
202
+ inprocess_barrier_timeout ....................... 120
203
+ inprocess_completion_timeout .................... 120
204
+ inprocess_empty_cuda_cache ...................... False
205
+ inprocess_granularity ........................... node
206
+ inprocess_hard_timeout .......................... 90
207
+ inprocess_heartbeat_interval .................... 30
208
+ inprocess_heartbeat_timeout ..................... 60
209
+ inprocess_last_call_wait ........................ 1
210
+ inprocess_max_iterations ........................ None
211
+ inprocess_monitor_process_interval .............. 1.0
212
+ inprocess_monitor_thread_interval ............... 1.0
213
+ inprocess_progress_watchdog_interval ............ 1.0
214
+ inprocess_restart ............................... False
215
+ inprocess_soft_timeout .......................... 60
216
+ inprocess_termination_grace_time ................ 1
217
+ is_hybrid_model ................................. False
218
+ iter_per_epoch .................................. 1250
219
+ iterations_to_skip .............................. []
220
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
221
+ kv_channels ..................................... 64
222
+ kv_lora_rank .................................... 32
223
+ lazy_mpu_init ................................... None
224
+ load ............................................ gpt-checkpoint
225
+ load_model_opt_format ........................... False
226
+ local_rank ...................................... 0
227
+ log_interval .................................... 1
228
+ log_loss_scale_to_tensorboard ................... True
229
+ log_memory_to_tensorboard ....................... False
230
+ log_num_zeros_in_grad ........................... False
231
+ log_params_norm ................................. False
232
+ log_progress .................................... False
233
+ log_straggler ................................... False
234
+ log_throughput .................................. False
235
+ log_timers_to_tensorboard ....................... False
236
+ log_validation_ppl_to_tensorboard ............... False
237
+ log_world_size_to_tensorboard ................... False
238
+ logging_level ................................... 0
239
+ loss_scale ...................................... None
240
+ loss_scale_window ............................... 1000
241
+ lr .............................................. 0.0005
242
+ lr_decay_iters .................................. 150000
243
+ lr_decay_samples ................................ None
244
+ lr_decay_style .................................. cosine
245
+ lr_warmup_fraction .............................. None
246
+ lr_warmup_init .................................. 0.0
247
+ lr_warmup_iters ................................. 2
248
+ lr_warmup_samples ............................... 0
249
+ lr_wsd_decay_iters .............................. None
250
+ lr_wsd_decay_samples ............................ None
251
+ lr_wsd_decay_style .............................. exponential
252
+ main_grads_dtype ................................ torch.float32
253
+ main_params_dtype ............................... torch.float32
254
+ make_vocab_size_divisible_by .................... 128
255
+ mamba_head_dim .................................. 64
256
+ mamba_num_groups ................................ 8
257
+ mamba_num_heads ................................. None
258
+ mamba_state_dim ................................. 128
259
+ manual_gc ....................................... False
260
+ manual_gc_eval .................................. True
261
+ manual_gc_interval .............................. 0
262
+ mask_factor ..................................... 1.0
263
+ mask_prob ....................................... 0.15
264
+ mask_type ....................................... random
265
+ masked_softmax_fusion ........................... True
266
+ max_position_embeddings ......................... 1024
267
+ max_tokens_to_oom ............................... 12000
268
+ memory_snapshot_path ............................ snapshot.pickle
269
+ merge_file ...................................... merges.txt
270
+ micro_batch_size ................................ 1
271
+ microbatch_group_size_per_vp_stage .............. None
272
+ mid_level_dataset_surplus ....................... 0.005
273
+ min_loss_scale .................................. 1.0
274
+ min_lr .......................................... 0.0
275
+ mlp_chunks_for_prefill .......................... 1
276
+ mmap_bin_files .................................. True
277
+ mock_data ....................................... True
278
+ moe_apply_probs_on_input ........................ False
279
+ moe_aux_loss_coeff .............................. 0.0
280
+ moe_enable_deepep ............................... False
281
+ moe_expert_capacity_factor ...................... None
282
+ moe_extended_tp ................................. False
283
+ moe_ffn_hidden_size ............................. None
284
+ moe_grouped_gemm ................................ False
285
+ moe_input_jitter_eps ............................ None
286
+ moe_layer_freq .................................. 1
287
+ moe_layer_recompute ............................. False
288
+ moe_pad_expert_input_to_capacity ................ False
289
+ moe_per_layer_logging ........................... False
290
+ moe_permute_fusion .............................. False
291
+ moe_router_bias_update_rate ..................... 0.001
292
+ moe_router_dtype ................................ None
293
+ moe_router_enable_expert_bias ................... False
294
+ moe_router_force_load_balancing ................. False
295
+ moe_router_group_topk ........................... None
296
+ moe_router_load_balancing_type .................. aux_loss
297
+ moe_router_num_groups ........................... None
298
+ moe_router_padding_for_fp8 ...................... False
299
+ moe_router_pre_softmax .......................... False
300
+ moe_router_score_function ....................... softmax
301
+ moe_router_topk ................................. 2
302
+ moe_router_topk_scaling_factor .................. None
303
+ moe_shared_expert_intermediate_size ............. None
304
+ moe_shared_expert_overlap ....................... False
305
+ moe_token_dispatcher_type ....................... allgather
306
+ moe_token_drop_policy ........................... probs
307
+ moe_use_legacy_grouped_gemm ..................... False
308
+ moe_use_upcycling ............................... False
309
+ moe_z_loss_coeff ................................ None
310
+ mrope_section ................................... None
311
+ mscale .......................................... 1.0
312
+ mscale_all_dim .................................. 1.0
313
+ mtp_loss_scaling_factor ......................... 0.1
314
+ mtp_num_layers .................................. None
315
+ multi_latent_attention .......................... False
316
+ nccl_all_reduce_for_prefill ..................... False
317
+ nccl_communicator_config_path ................... None
318
+ nccl_ub ......................................... False
319
+ no_load_optim ................................... None
320
+ no_load_rng ..................................... None
321
+ no_persist_layer_norm ........................... False
322
+ no_rope_freq .................................... None
323
+ no_save_optim ................................... None
324
+ no_save_rng ..................................... None
325
+ non_persistent_ckpt_type ........................ None
326
+ non_persistent_global_ckpt_dir .................. None
327
+ non_persistent_local_ckpt_algo .................. fully_parallel
328
+ non_persistent_local_ckpt_dir ................... None
329
+ non_persistent_save_interval .................... None
330
+ norm_epsilon .................................... 1e-05
331
+ normalization ................................... LayerNorm
332
+ num_attention_heads ............................. 64
333
+ num_channels .................................... 3
334
+ num_classes ..................................... 1000
335
+ num_dataset_builder_threads ..................... 1
336
+ num_distributed_optimizer_instances ............. 1
337
+ num_experts ..................................... None
338
+ num_layers ...................................... 2
339
+ num_layers_at_end_in_bf16 ....................... 1
340
+ num_layers_at_start_in_bf16 ..................... 1
341
+ num_layers_per_virtual_pipeline_stage ........... None
342
+ num_query_groups ................................ 16
343
+ num_virtual_stages_per_pipeline_rank ............ None
344
+ num_workers ..................................... 2
345
+ object_storage_cache_path ....................... None
346
+ one_logger_async ................................ False
347
+ one_logger_project .............................. megatron-lm
348
+ one_logger_run_name ............................. None
349
+ onnx_safe ....................................... None
350
+ openai_gelu ..................................... False
351
+ optimizer ....................................... adam
352
+ optimizer_cpu_offload ........................... False
353
+ optimizer_offload_fraction ...................... 1.0
354
+ output_bert_embeddings .......................... False
355
+ overlap_cpu_optimizer_d2h_h2d ................... False
356
+ overlap_grad_reduce ............................. False
357
+ overlap_p2p_comm ................................ False
358
+ overlap_p2p_comm_warmup_flush ................... False
359
+ overlap_param_gather ............................ False
360
+ overlap_param_gather_with_optimizer_step ........ False
361
+ override_opt_param_scheduler .................... False
362
+ params_dtype .................................... torch.float16
363
+ patch_dim ....................................... 16
364
+ per_split_data_args_path ........................ None
365
+ perform_initialization .......................... True
366
+ pin_cpu_grads ................................... True
367
+ pin_cpu_params .................................. True
368
+ pipeline_model_parallel_comm_backend ............ None
369
+ pipeline_model_parallel_size .................... 1
370
+ pipeline_model_parallel_split_rank .............. None
371
+ position_embedding_type ......................... learned_absolute
372
+ pretrained_checkpoint ........................... None
373
+ profile ......................................... False
374
+ profile_ranks ................................... [0]
375
+ profile_step_end ................................ 12
376
+ profile_step_start .............................. 10
377
+ q_lora_rank ..................................... None
378
+ qk_head_dim ..................................... 128
379
+ qk_l2_norm ...................................... False
380
+ qk_layernorm .................................... False
381
+ qk_pos_emb_head_dim ............................. 64
382
+ query_in_block_prob ............................. 0.1
383
+ rampup_batch_size ............................... None
384
+ rank ............................................ 0
385
+ recompute_granularity ........................... None
386
+ recompute_method ................................ None
387
+ recompute_modules ............................... None
388
+ recompute_num_layers ............................ None
389
+ record_memory_history ........................... False
390
+ relative_attention_max_distance ................. 128
391
+ relative_attention_num_buckets .................. 32
392
+ replication ..................................... False
393
+ replication_factor .............................. 2
394
+ replication_jump ................................ None
395
+ rerun_mode ...................................... disabled
396
+ reset_attention_mask ............................ False
397
+ reset_position_ids .............................. False
398
+ result_rejected_tracker_filename ................ None
399
+ retriever_report_topk_accuracies ................ []
400
+ retriever_score_scaling ......................... False
401
+ retriever_seq_length ............................ 256
402
+ retro_add_retriever ............................. False
403
+ retro_attention_gate ............................ 1
404
+ retro_cyclic_train_iters ........................ None
405
+ retro_encoder_attention_dropout ................. 0.1
406
+ retro_encoder_hidden_dropout .................... 0.1
407
+ retro_encoder_layers ............................ 2
408
+ retro_num_neighbors ............................. 2
409
+ retro_num_retrieved_chunks ...................... 2
410
+ retro_project_dir ............................... None
411
+ retro_verify_neighbor_count ..................... True
412
+ rope_scaling_factor ............................. 8.0
413
+ rotary_base ..................................... 10000
414
+ rotary_interleaved .............................. False
415
+ rotary_percent .................................. 1.0
416
+ rotary_scaling_factor ........................... 1.0
417
+ rotary_seq_len_interpolation_factor ............. None
418
+ run_workload_inspector_server ................... False
419
+ sample_rate ..................................... 1.0
420
+ save ............................................ gpt-checkpoint
421
+ save_interval ................................... 16
422
+ scatter_gather_tensors_in_pipeline .............. True
423
+ seed ............................................ 1234
424
+ seq_length ...................................... 1024
425
+ sequence_parallel ............................... False
426
+ sgd_momentum .................................... 0.9
427
+ short_seq_prob .................................. 0.1
428
+ skip_train ...................................... False
429
+ skipped_train_samples ........................... 0
430
+ spec ............................................ None
431
+ split ........................................... None
432
+ squared_relu .................................... False
433
+ start_weight_decay .............................. 0.1
434
+ straggler_ctrlr_port ............................ 65535
435
+ straggler_minmax_count .......................... 1
436
+ suggested_communication_unit_size ............... None
437
+ swiglu .......................................... False
438
+ swin_backbone_type .............................. tiny
439
+ symmetric_ar_type ............................... None
440
+ te_rng_tracker .................................. False
441
+ tensor_model_parallel_size ...................... 8
442
+ tensorboard_dir ................................. tensorboard-logs/
443
+ tensorboard_log_interval ........................ 1
444
+ tensorboard_queue_size .......................... 1000
445
+ test_data_path .................................. None
446
+ test_mode ....................................... False
447
+ tiktoken_num_special_tokens ..................... 1000
448
+ tiktoken_pattern ................................ None
449
+ tiktoken_special_tokens ......................... None
450
+ timing_log_level ................................ 0
451
+ timing_log_option ............................... minmax
452
+ titles_data_path ................................ None
453
+ tokenizer_model ................................. None
454
+ tokenizer_type .................................. GPT2BPETokenizer
455
+ torch_fsdp2_reshard_after_forward ............... True
456
+ tp_comm_bootstrap_backend ....................... nccl
457
+ tp_comm_bulk_dgrad .............................. True
458
+ tp_comm_bulk_wgrad .............................. True
459
+ tp_comm_overlap ................................. False
460
+ tp_comm_overlap_ag .............................. True
461
+ tp_comm_overlap_cfg ............................. None
462
+ tp_comm_overlap_rs .............................. True
463
+ tp_comm_overlap_rs_dgrad ........................ False
464
+ tp_comm_split_ag ................................ True
465
+ tp_comm_split_rs ................................ True
466
+ train_data_path ................................. None
467
+ train_iters ..................................... 10
468
+ train_samples ................................... None
469
+ train_sync_interval ............................. None
470
+ transformer_impl ................................ transformer_engine
471
+ transformer_pipeline_model_parallel_size ........ 1
472
+ untie_embeddings_and_output_weights ............. False
473
+ use_checkpoint_args ............................. False
474
+ use_checkpoint_opt_param_scheduler .............. False
475
+ use_cpu_initialization .......................... None
476
+ use_custom_fsdp ................................. False
477
+ use_dist_ckpt ................................... True
478
+ use_dist_ckpt_deprecated ........................ False
479
+ use_distributed_optimizer ....................... False
480
+ use_flash_attn .................................. False
481
+ use_legacy_models ............................... False
482
+ use_mp_args_from_checkpoint_args ................ False
483
+ use_one_sent_docs ............................... False
484
+ use_persistent_ckpt_worker ...................... False
485
+ use_precision_aware_optimizer ................... False
486
+ use_pytorch_profiler ............................ False
487
+ use_ring_exchange_p2p ........................... False
488
+ use_rope_scaling ................................ False
489
+ use_rotary_position_embeddings .................. False
490
+ use_sharp ....................................... False
491
+ use_tokenizer_model_from_checkpoint_args ........ True
492
+ use_torch_fsdp2 ................................. False
493
+ use_torch_optimizer_for_cpu_offload ............. False
494
+ use_tp_pp_dp_mapping ............................ False
495
+ v_head_dim ...................................... 128
496
+ valid_data_path ................................. None
497
+ variable_seq_lengths ............................ False
498
+ virtual_pipeline_model_parallel_size ............ None
499
+ vision_backbone_type ............................ vit
500
+ vision_pretraining .............................. False
501
+ vision_pretraining_type ......................... classify
502
+ vocab_extra_ids ................................. 0
503
+ vocab_file ...................................... vocab.json
504
+ vocab_size ...................................... None
505
+ wandb_exp_name ..................................
506
+ wandb_project ...................................
507
+ wandb_save_dir ..................................
508
+ weight_decay .................................... 0.1
509
+ weight_decay_incr_style ......................... constant
510
+ wgrad_deferral_limit ............................ 0
511
+ world_size ...................................... 8
512
+ yaml_cfg ........................................ None
513
+ -------------------- end of arguments ---------------------
514
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
515
+ > building GPT2BPETokenizer tokenizer ...
516
+ INFO:megatron.training.initialize:Setting logging level to 0
517
+ WARNING: TensorBoard writing requested but is not available (are you using PyTorch 1.1.0 or later?), no TensorBoard logs will be written.
518
+ WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it
519
+ INFO:megatron.training.initialize:Setting logging level to 0
520
+ > padded vocab (size: 50257) with 943 dummy tokens (new size: 51200)
521
+ INFO:megatron.training.initialize:Setting logging level to 0
522
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.DISABLED
523
+ > initializing torch distributed ...
524
+ > initialized tensor model parallel with size 8
525
+ > initialized pipeline model parallel with size 1
526
+ > setting random seeds to 1234 ...
527
+ > compiling dataset index builder ...
528
+ make: Entering directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets'
529
+ INFO:megatron.training.initialize:Setting logging level to 0
530
+ INFO:megatron.training.initialize:Setting logging level to 0
531
+ INFO:megatron.training.initialize:Setting logging level to 0
532
+ INFO:megatron.training.initialize:Setting logging level to 0
533
+ make: Nothing to be done for 'default'.
534
+ make: Leaving directory '/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/datasets'
535
+ >>> done with dataset index builder. Compilation time: 0.047 seconds
536
+ > compiling and loading fused kernels ...