Commits

Breadcrumbs

History for

onrefs/heads/main

User selector

All users

Datepicker

All time

Commit History

Commits on Aug 27, 2024

• [Bug] Fix crash when logging optimizer state to tb (#417)

  billishyahao

  authored
  1280f59

Commits on Aug 26, 2024

• enable empty cache on XPU device (#438)

  ys950902

  authored
  8e9d973

Commits on Aug 24, 2024

• fix nan issue when running megatron-deepspeed (#434)

  ys950902

  authored
  4f9f1f6

Commits on Jul 17, 2024

• add kill switch file support to gracefully exit training at runtime (#412)

  Show description for fc989b8
  polisettyvarma

  authored
  fc989b8

Commits on Jul 16, 2024

• [Wandb] Refine wandb logging function (#416)

  Show description for 73029ed
  billishyahao

  authored
  73029ed

Commits on Jul 15, 2024

• add PyTorch profiler support (#414)

  Show description for 0971e68
  polisettyvarma

  authored
  0971e68

Commits on May 13, 2024

• get distributed backend name via accelerator and check loss_scale before writing to tb (#374)

  Show description for 7eb36a1
  polisettyvarma

  authored
  7eb36a1

Commits on Apr 9, 2024

• Support MoE for GPTModelPipe (#373)

  Show description for bcedecd
  mosheisland

  authored
  bcedecd

Commits on Feb 21, 2024

• Track additional metrics with W&B in megatron/training.py (#348)

  Show description for c934137
  saforem2

  authored
  c934137

Commits on Feb 1, 2024

• Track additional metrics with W&B in megatron/training.py (#343)

  saforem2

  authored
  6d4c535

Commits on Jan 5, 2024

• Revert "Modify the args_default usage" (#325)

  Show description for d65921c
  yuanwu2017

  authored
  d65921c

Commits on Nov 14, 2023

• Add tgs metrics (#286)

  CokeDong

  authored
  155ce98

• Enable the args.deepspeed_config to use dict type (#290)

  Show description for 15355af
  yuanwu2017

  authored
  15355af

Commits on Nov 9, 2023

• Enable the combination of sequence length warmup and RoPE (#285)

  conglongli

  authored
  4c0bc7f

Commits on Oct 24, 2023

• fix the config file unclose (#268)

  Show description for 796866f
  zhangsheng377

  authored
  796866f

Commits on Oct 23, 2023

• Enable universal checkpointing (#265)

  Show description for 54e6c2c
  

  tjruwase

  and

  mrwyattii

  authored
  54e6c2c

Commits on Sep 21, 2023

• Add support for logging throughput with W&B (#256)

  Show description for 3c3d4cf
  saforem2

  authored
  3c3d4cf

Commits on Aug 23, 2023

• DeepSpeed Ulysses Release (#230)

  Show description for a2a476e
  

  3 peopleauthored
  a2a476e

Commits on Jul 18, 2023

• fix error comment in megatron/training.py (#165)

  YizhouZ

  authored
  55d322d

• support other devices besides cuda (#162)

  Show description for 8c0fd47
  YizhouZ

  authored
  8c0fd47

Commits on Jul 12, 2023

• refactor and fixes

  conglongli

  committed
  636540a

Commits on Jul 4, 2023

• Merge remote-tracking branch 'upstream/main' into rebase

  conglongli

  committed
  9e72d57

Commits on Jun 28, 2023

• Merge branch 'main' into core_transformer

  jaredcasper

  committed
  e68e57a

Commits on Jun 23, 2023

• Give CLI option to skip the training loop

  

  John Kamalu

  committed
  a82739c

Commits on Jun 13, 2023

• Retro fix for tensor parallelism.

  

  lmcafee-nvidia

  authored and

  jaredcasper

  committed
  f479999

Commits on May 31, 2023

• Consistent arg names.

  jaredcasper

  committed
  2c13d1f

• Add option to overlap p2p communication.

  jaredcasper

  committed
  f9283c5

Commits on May 30, 2023

• Merge branch 'main' into core_transformer

  jaredcasper

  committed
  f786fab

• Fix evaluatin in training.py with new config objects.

  jaredcasper

  committed
  e8c06f4

Commits on May 26, 2023

• Retro updates.

  

  lmcafee-nvidia

  authored and

  jaredcasper

  committed
  6dba84b

Commits on May 25, 2023

• Another rework of pipeline arguments/configuration

  Show description for 9d83398
  jaredcasper

  committed
  9d83398

Commits on May 19, 2023

• Split pipeline config into separate object and various fixes.

  jaredcasper

  committed
  31d133b

Commits on May 18, 2023

• Split TransformerConfig into BaseConfig and TransformerConfig, use BaseConfig for model parallel functions.

  Show description for 397d0b2
  jaredcasper

  committed
  397d0b2

Commits on May 10, 2023

• fix(training.py): logical bug in eval_iters_calculation (#138)

  Show description for 1f640c0
  2catycm

  authored
  1f640c0

Commits on May 6, 2023

• Add a space to fix tab error for loss_scale (#136)

  Show description for ec70e81
  ys950902

  authored
  ec70e81

Pagination

Previous

Next