Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit History

Commits on May 2, 2024

• Merge branch 'boxiangw/mlperf-option-add-one-extra-token' into 'main'

  Show description for db3a3f7
  ericharper

  committed
  db3a3f7

  Copy full SHA for db3a3f7

• [MLPerf] GPT dataset features: drop last partial validation sequence, drop extra token, return sample with 1s loss mask, mock dataset testing

  

  BoxiangW

  authored and

  ericharper

  committed
  c90aa16

  Copy full SHA for c90aa16

• Merge branch 'dnarayanan/remove_buckets_from_checkpoints' into 'main'

  Show description for 2297178
  shanmugamr1992

  committed
  2297178

  Copy full SHA for 2297178

• Merge branch 'malay/fwd_step_fix' into 'main'

  Show description for ec16db0
  shanmugamr1992

  committed
  ec16db0

  Copy full SHA for ec16db0

• Merge branch '147-optimize-the-hidden-dimension-ag-rs-implementation' into 'main'

  Show description for 72560a8
  shanmugamr1992

  committed
  72560a8

  Copy full SHA for 72560a8

• Optimize the Hidden Dimension AG/RS implementation

  

  yanring

  authored and

  shanmugamr1992

  committed
  a8550c1

  Copy full SHA for a8550c1

• [MoE] Token Drop Support

  

  yanring

  authored and

  ericharper

  committed
  e2ec14a

  Copy full SHA for e2ec14a

Commits on May 1, 2024

• Extended TP for MOE + memory checkpointing

  

  kvareddy

  authored and

  jaredcasper

  committed
  e61c384

  Copy full SHA for e61c384

Commits on Apr 30, 2024

• Merge branch 'mblaz/fix-qk-layernorm-test' into 'main'

  Show description for 57e1262
  ericharper

  committed
  57e1262

  Copy full SHA for 57e1262

• Fix quotes in qk-layernorm test

  

  mikolajblaz

  authored and

  ericharper

  committed
  10ceb4d

  Copy full SHA for 10ceb4d

• Make distributed optimizer checkpoint agnostic to the bucket size

  deepakn94

  committed
  369e698

  Copy full SHA for 369e698

Commits on Apr 29, 2024

• Merge branch 'trintamaki/linear-tp' into 'main'

  Show description for 20674cc
  jaredcasper

  committed
  20674cc

  Copy full SHA for 20674cc

• LinearWithFrozenWeight backward fix when TP > 1

  

  trintamaki

  authored and

  jaredcasper

  committed
  5fffdfc

  Copy full SHA for 5fffdfc

• Merge branch 'loss_averaging' into 'main'

  Show description for 0264800
  jaredcasper

  committed
  0264800

  Copy full SHA for 0264800

• Fix Cross Entropy Loss Averaging

  

  Mike Chrzanowski

  authored and

  jaredcasper

  committed
  adfa873

  Copy full SHA for adfa873

Commits on Apr 26, 2024

• Merge branch 'main' into mblaz/fpsl-model-weights

  mikolajblaz

  committed
  f297495

  Copy full SHA for f297495

• Compute hashes on each rank, and compare across DP replicas

  

  deepakn94

  authored and

  ericharper

  committed
  2afccb6

  Copy full SHA for 2afccb6

Commits on Apr 22, 2024

• Update minor version

  ericharper

  committed
  f0d3220

  Copy full SHA for f0d3220

Commits on Apr 19, 2024

• Fix formatting

  deepakn94

  committed
  bbf6613

  Copy full SHA for bbf6613

• forward step missing arg

  

  Malay Nagda

  committed
  2dbcaeb

  Copy full SHA for 2dbcaeb

• [TMP] Verbose logging

  mikolajblaz

  committed
  47ae952

  Copy full SHA for 47ae952

• Merge branch 'mblaz/add-ckpt-resume-functional-tests' into mblaz/fpsl-model-weights-load

  mikolajblaz

  committed
  61f3a9a

  Copy full SHA for 61f3a9a

Commits on Apr 18, 2024

• fix EP distopt with overlap param gather

  

  shjwudp

  authored and

  ericharper

  committed
  3833a0e

  Copy full SHA for 3833a0e

• Fix dtype grouping

  mikolajblaz

  committed
  f714ff1

  Copy full SHA for f714ff1

• Fix args usage

  mikolajblaz

  committed
  805caac

  Copy full SHA for 805caac

• Merge branch 'mblaz/fpsl-model-weights' into mblaz/fpsl-model-weights-load

  mikolajblaz

  committed
  2369270

  Copy full SHA for 2369270

• Merge branch 'main' into mblaz/fpsl-model-weights

  mikolajblaz

  committed
  68eac5c

  Copy full SHA for 68eac5c

• Merge branch 'mblaz/fpsl-model-weights' into mblaz/fpsl-model-weights-load

  mikolajblaz

  committed
  48efe4b

  Copy full SHA for 48efe4b

• Added fusion for squared relu

  

  sanandaraj5597

  authored and

  jaredcasper

  committed
  d06c76c

  Copy full SHA for d06c76c

Commits on Apr 17, 2024

• Support missing size and missing weights in dataset creation and blending

  

  John Kamalu

  authored and

  ericharper

  committed
  dc52e84

  Copy full SHA for dc52e84

• Merge branch 'pfischer/re-fix-pip-import' into 'main'

  Show description for ad354e0
  jaredcasper

  committed
  ad354e0

  Copy full SHA for ad354e0

• Revert lmcafee/pip-import-fix and fix setup.py instead

  

  philipp-fischer

  authored and

  jaredcasper

  committed
  0d982fc

  Copy full SHA for 0d982fc

• ci hack

  

  wdykas

  authored and

  shanmugamr1992

  committed
  cda03f5

  Copy full SHA for cda03f5

• Add _ShardId doc

  mikolajblaz

  committed
  00a6a3a

  Copy full SHA for 00a6a3a

• Fix formatting

  mikolajblaz

  committed
  d1a9e24

  Copy full SHA for d1a9e24

Pagination

Previous

Next