Code

Add perf model configs gb200#12140

Code

Merged

erhoo82 merged 13 commits intoNVIDIA:mainNVIDIA/NeMo:mainfrom guyueh1:add_perf_model_configs_gb200guyueh1/NeMo:add_perf_model_configs_gb200Copy head branch name to clipboardFeb 14, 2025

+71-47

Commits

Commits on Feb 10, 2025

• Bug fix with generation of expert_tensor_parallel_rank

  Show description for 8cc23e8
  guyueh1

  committed
  8cc23e8

  Copy full SHA for 8cc23e8

• Fix pylint

  Show description for bc8441e
  guyueh1

  committed
  bc8441e

  Copy full SHA for bc8441e

Commits on Feb 11, 2025

• Support perf script with gb200

  Show description for 37dbae3
  guyueh1

  committed
  37dbae3

  Copy full SHA for 37dbae3

• fix

  Show description for 2db23e2
  guyueh1

  committed
  2db23e2

  Copy full SHA for 2db23e2

• mixtral 8x22b recipe change to mitigate a megatron.core bug

  Show description for 8d87cbf
  guyueh1

  committed
  8d87cbf

  Copy full SHA for 8d87cbf

Commits on Feb 12, 2025

• Merge branch 'main' into add_perf_model_configs_gb200

  guyueh1

  authored
  135d542

  Copy full SHA for 135d542

• Make column names of all csvs align; fix for 8x22b

  Show description for 215b51e
  guyueh1

  committed
  215b51e

  Copy full SHA for 215b51e

• typing fix

  Show description for 3071d05
  guyueh1

  committed
  3071d05

  Copy full SHA for 3071d05

• Merge branch 'add_perf_model_configs_gb200' of github.com:guyueh1/NeMo into add_perf_model_configs_gb200

  guyueh1

  committed
  bbb3947

  Copy full SHA for bbb3947

• Adjust 405B config due to asymmetric PP

  Show description for a953197
  guyueh1

  committed
  a953197

  Copy full SHA for a953197

• Add a comment

  Show description for 7e51eab
  guyueh1

  committed
  7e51eab

  Copy full SHA for 7e51eab

• Change 405B GBS

  Show description for 0f28241
  guyueh1

  committed
  0f28241

  Copy full SHA for 0f28241

• fix gbs for gb200 finetuning

  Show description for 18fa9d4
  guyueh1

  committed
  18fa9d4

  Copy full SHA for 18fa9d4