Code

Upstream sync#6

Code

Merged

gurpreet-dhami merged 147 commits intorocm_devROCm/Megatron-LM:rocm_devfrom upstream_syncROCm/Megatron-LM:upstream_syncCopy head branch name to clipboardOct 18, 2024

+10,369-2,275

Commits

Commits on Sep 11, 2024

• ADLR/megatron-lm!2092 - ci: Bump reference sha

  ko3n1g

  committed
  2130890

  Copy full SHA for 2130890

• Merge branch 'ko3n1g/ci/bump-sha-3' into 'main'

  Show description for 6664dc6
  ko3n1g

  committed
  6664dc6

  Copy full SHA for 6664dc6

• ADLR/megatron-lm!2093 - ci: Disable broken test

  ko3n1g

  committed
  32949f2

  Copy full SHA for 32949f2

• Merge branch 'ko3n1g/ci/disable-broken-test' into 'main'

  Show description for df1418a
  ko3n1g

  committed
  df1418a

  Copy full SHA for df1418a

• ADLR/megatron-lm!1985 - Multimodal sequence length optimizations

  

  trintamaki

  authored and

  jaredcasper

  committed
  f8b7c3f

  Copy full SHA for f8b7c3f

• Merge branch 'trintamaki/multi-image-multi-tile-dataloader-seq-len' into 'main'

  Show description for 6151709
  jaredcasper

  committed
  6151709

  Copy full SHA for 6151709

• ADLR/megatron-lm!2094 - tests: Disable flaky test

  ko3n1g

  committed
  3005d02

  Copy full SHA for 3005d02

• Merge branch 'ko3n1g/tests/flaky-test-2' into 'main'

  Show description for 9ec2337
  ko3n1g

  committed
  9ec2337

  Copy full SHA for 9ec2337

Commits on Sep 12, 2024

• ADLR/megatron-lm!2004 - tests: Repeat MRs 5 times

  ko3n1g

  committed
  e5fb1fa

  Copy full SHA for e5fb1fa

• Merge branch 'ko3n1g/ci/repeat-mrs' into 'main'

  Show description for 028b777
  ko3n1g

  committed
  028b777

  Copy full SHA for 028b777

• ADLR/megatron-lm!2091 - Don't pass device_id to torch.distributed.init_process_group, it causes hangs

  Show description for dcc6634
  

  szmigacz

  authored and

  ko3n1g

  committed
  dcc6634

  Copy full SHA for dcc6634

• Merge branch 'no_dist_device_id' into 'main'

  Show description for 76f9f48
  ko3n1g

  committed
  76f9f48

  Copy full SHA for 76f9f48

Commits on Sep 14, 2024

• ADLR/megatron-lm!2059 - ci: Add release tests for 0.9

  ko3n1g

  committed
  bf7b978

  Copy full SHA for bf7b978

• Merge branch 'ko3n1g/ci/release-tests' into 'main'

  Show description for 21924d8
  ko3n1g

  committed
  21924d8

  Copy full SHA for 21924d8

Commits on Sep 17, 2024

• ADLR/megatron-lm!2106 - fix: allow merge request CI for non-protected branches to fail

  

  terrykong

  authored and

  ko3n1g

  committed
  e6f1d81

  Copy full SHA for e6f1d81

• Merge branch 'terryk/ci-can-fail-on-unprotected-targets' into 'main'

  Show description for 6562666
  ko3n1g

  committed
  6562666

  Copy full SHA for 6562666

• ADLR/megatron-lm!2107 - chore: Fix autoformatter for release branches

  ko3n1g

  committed
  0902af0

  Copy full SHA for 0902af0

• Merge branch 'ko3n1g/chore/formatting-on-release-branch' into 'main'

  Show description for 72008a0
  ko3n1g

  committed
  72008a0

  Copy full SHA for 72008a0

• ADLR/megatron-lm!2104 - Fixing broken links

  Show description for 2a8d8af
  

  shanmugamr1992

  and

  Shanmugam Ramasamy

  committed
  2a8d8af

  Copy full SHA for 2a8d8af

• Merge branch 'docFix' into 'main'

  Show description for 3f10ff6
  shanmugamr1992

  committed
  3f10ff6

  Copy full SHA for 3f10ff6

• ADLR/megatron-lm!2072 - Add video handling into multimodal mcore

  

  Matthieu Le

  authored and

  jon-barker

  committed
  71d8ce7

  Copy full SHA for 71d8ce7

• Merge branch 'add-video-handling' into 'main'

  Show description for 0bda578
  jon-barker

  committed
  0bda578

  Copy full SHA for 0bda578

Commits on Sep 18, 2024

• ADLR/megatron-lm!1715 - Enable optional kwargs with CUDA graph

  

  vasunvidia

  authored and

  ko3n1g

  committed
  ab7f706

  Copy full SHA for ab7f706

• Merge branch 'lora_cg' into 'main'

  Show description for 77b4bfe
  ko3n1g

  committed
  77b4bfe

  Copy full SHA for 77b4bfe

• ADLR/megatron-lm!2077 - Resolve "Fix TE version in TELinear"

  

  Victarry

  authored and

  ericharper

  committed
  0cffc6b

  Copy full SHA for 0cffc6b

• Merge branch '318-fix-te-version-in-telinear' into 'main'

  Show description for 461b06c
  ericharper

  committed
  461b06c

  Copy full SHA for 461b06c

• ADLR/megatron-lm!2112 - Update path to MMMU to use new repos structure

  

  Matthieu Le

  authored and

  jon-barker

  committed
  6b78cb1

  Copy full SHA for 6b78cb1

• Merge branch 'fix_mmmu_mmodal' into 'main'

  Show description for d350231
  jon-barker

  committed
  d350231

  Copy full SHA for d350231

• ADLR/megatron-lm!1880 - Removing env variable NVTE_ALLOW_NONDETERMINISTIC_ALGO

  Show description for cedd415
  

  shanmugamr1992

  and

  Shanmugam Ramasamy

  committed
  cedd415

  Copy full SHA for cedd415

• Merge branch 'bertflash' into 'main'

  Show description for 6b35ca8
  shanmugamr1992

  committed
  6b35ca8

  Copy full SHA for 6b35ca8

Commits on Sep 19, 2024

• ADLR/megatron-lm!2033 - Online eval

  

  trintamaki

  authored and

  ericharper

  committed
  63be779

  Copy full SHA for 63be779

• Merge branch 'trintamaki/online-eval' into 'main'

  Show description for 835af44
  ericharper

  committed
  835af44

  Copy full SHA for 835af44

• ADLR/megatron-lm!1973 - MMMU multi-image support

  

  trintamaki

  authored and

  jon-barker

  committed
  2c9bcac

  Copy full SHA for 2c9bcac

• Merge branch 'trintamaki/multi-image-mmmu' into 'main'

  Show description for 905de33
  jon-barker

  committed
  905de33

  Copy full SHA for 905de33

Commits on Sep 20, 2024

• ADLR/megatron-lm!2113 - build: Use multi-stage for parallel builds

  ko3n1g

  committed
  5c0697c

  Copy full SHA for 5c0697c

• Merge branch 'ko3n1g/build/pip' into 'main'

  Show description for c394f78
  ko3n1g

  committed
  c394f78

  Copy full SHA for c394f78

Commits on Sep 21, 2024

• ADLR/megatron-lm!2126 - Only print warning when relevant

  

  deepakn94

  authored and

  jaredcasper

  committed
  cf596b9

  Copy full SHA for cf596b9

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

  Show description for 640e62f
  jaredcasper

  committed
  640e62f

  Copy full SHA for 640e62f

• ADLR/megatron-lm!2124 - tests: Fix location of megatron

  ko3n1g

  committed
  3eeb932

  Copy full SHA for 3eeb932

• Merge branch 'ko3n1g/tests/fix-location-of-megatron' into 'main'

  Show description for 205f946
  ko3n1g

  committed
  205f946

  Copy full SHA for 205f946

• ADLR/megatron-lm!2127 - ci: Bump sha

  ko3n1g

  committed
  d210eb0

  Copy full SHA for d210eb0

• Merge branch 'ko3n1g/chore/bump-sha' into 'main'

  Show description for 811a26a
  ko3n1g

  committed
  811a26a

  Copy full SHA for 811a26a

Commits on Sep 22, 2024

• ADLR/megatron-lm!2128 - ci: Improve cherry pick workflow

  ko3n1g

  committed
  405135a

  Copy full SHA for 405135a

• Merge branch 'ko3n1g/ci/improve-cherry-pick-workflow' into 'main'

  Show description for fba615f
  ko3n1g

  committed
  fba615f

  Copy full SHA for fba615f

• ADLR/megatron-lm!2034 - ci: Introduce JET Python SDK

  ko3n1g

  committed
  95be3cb

  Copy full SHA for 95be3cb

• Merge branch 'ko3n1g/ci/convergence-tests-with-jet' into 'main'

  Show description for e79808c
  ko3n1g

  committed
  e79808c

  Copy full SHA for e79808c

• ADLR/megatron-lm!2130 - ci: Improve cherry pick MR description

  ko3n1g

  committed
  e10a9f4

  Copy full SHA for e10a9f4

• Merge branch 'ko3n1g/ci/improve-cherry-pick-workflow' into 'main'

  Show description for 8e69382
  ko3n1g

  committed
  8e69382

  Copy full SHA for 8e69382

Commits on Sep 23, 2024

• ADLR/megatron-lm!2119 - Huvu/t5 te10 fix nemoci pr482

  Show description for e35818d
  

  2 peopleauthored and

  ericharper

  committed
  e35818d

  Copy full SHA for e35818d

• Merge branch 'huvu/t5_TE10_fix_nemoci_PR482' into 'main'

  Show description for dbd2d18
  ericharper

  committed
  dbd2d18

  Copy full SHA for dbd2d18

• ADLR/megatron-lm!2134 - ci: Set author and milestone for cherry-picks

  ko3n1g

  committed
  8c666c2

  Copy full SHA for 8c666c2

• Merge branch 'ko3n1g/ci/cherry-pick-authro' into 'main'

  Show description for 6d8dc80
  ko3n1g

  committed
  6d8dc80

  Copy full SHA for 6d8dc80

• ADLR/megatron-lm!2135 - ci: Send alerts on unit-tests-extended

  ko3n1g

  committed
  c45f951

  Copy full SHA for c45f951

• Merge branch 'ko3n1g/ci/notify-ut' into 'main'

  Show description for 08e80b0
  ko3n1g

  committed
  08e80b0

  Copy full SHA for 08e80b0

• ADLR/megatron-lm!2133 - tests: Minor improvements to JET

  ko3n1g

  committed
  643e60a

  Copy full SHA for 643e60a

• Merge branch 'ko3n1g/ci/fixes-to-jet' into 'main'

  Show description for 8ec4617
  ko3n1g

  committed
  8ec4617

  Copy full SHA for 8ec4617

• ADLR/megatron-lm!2136 - tests: Fix GPT test

  ko3n1g

  committed
  5ade91a

  Copy full SHA for 5ade91a

• Merge branch 'ko3n1g/tests/fix-gpt-release-samples' into 'main'

  Show description for 1f2d556
  ko3n1g

  committed
  1f2d556

  Copy full SHA for 1f2d556

• ADLR/megatron-lm!2139 - ci: Fix cherry-pick strings

  ko3n1g

  committed
  e464e94

  Copy full SHA for e464e94

• Merge branch 'ko3n1g/ci/cherry-pick-strip-chars' into 'main'

  Show description for 0fd4617
  ko3n1g

  committed
  0fd4617

  Copy full SHA for 0fd4617

• ADLR/megatron-lm!2110 - Use torch dataloader in multimodal evaluation

  

  trintamaki

  authored and

  jon-barker

  committed
  ede39b8

  Copy full SHA for ede39b8

• Merge branch 'trintamaki/multimodal-eval-dataset' into 'main'

  Show description for 2065c35
  jon-barker

  committed
  2065c35

  Copy full SHA for 2065c35

• ADLR/megatron-lm!2137 - ci: Enable dev container for new features

  ko3n1g

  committed
  697ea61

  Copy full SHA for 697ea61

• Merge branch 'ko3n1g/ci/dev-container' into 'main'

  Show description for 075c727
  ko3n1g

  committed
  075c727

  Copy full SHA for 075c727

Commits on Sep 24, 2024

• ADLR/megatron-lm!2005 - Fix performance regression brought by torch.bincount

  

  xxuwenc

  authored and

  ko3n1g

  committed
  5e23e72

  Copy full SHA for 5e23e72

• Merge branch 'revert_bincount' into 'main'

  Show description for 884b087
  ko3n1g

  committed
  884b087

  Copy full SHA for 884b087

• ADLR/megatron-lm!2073 - Multimodal batched bug fix

  

  trintamaki

  authored and

  jon-barker

  committed
  ad38459

  Copy full SHA for ad38459

• Merge branch 'trintamaki/multimodal_batch_bugfix' into 'main'

  Show description for 162b82d
  jon-barker

  committed
  162b82d

  Copy full SHA for 162b82d

• ADLR/megatron-lm!1581 - Add MLA support into MCore

  Show description for 32eac88
  

  2 peopleauthored and

  jaredcasper

  committed
  32eac88

  Copy full SHA for 32eac88

• Merge branch 'boxiangw/mla' into 'main'

  Show description for dcf9e77
  jaredcasper

  committed
  dcf9e77

  Copy full SHA for dcf9e77

Commits on Sep 25, 2024

• ADLR/megatron-lm!1995 - Add freeze options to pretrain_vlm

  

  trintamaki

  authored and

  jon-barker

  committed
  d207755

  Copy full SHA for d207755

• Merge branch 'trintamaki/pretrain_vlm_freeze_option' into 'main'

  Show description for 891b8f9
  jon-barker

  committed
  891b8f9

  Copy full SHA for 891b8f9

• ADLR/megatron-lm!2145 - Improve logging when decreasing batch size

  

  deepakn94

  authored and

  ericharper

  committed
  31c23f5

  Copy full SHA for 31c23f5

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

  Show description for 78bef1c
  ericharper

  committed
  78bef1c

  Copy full SHA for 78bef1c

• ADLR/megatron-lm!2148 - Add model.eval() to run_text_generation_server.py

  

  mathemakitten

  authored and

  jaredcasper

  committed
  5aceacb

  Copy full SHA for 5aceacb

• Merge branch 'hn-set-model-eval-mode' into 'main'

  Show description for 4158084
  jaredcasper

  committed
  4158084

  Copy full SHA for 4158084

Commits on Sep 26, 2024

• ADLR/megatron-lm!2111 - Mcore llama3.1 support

  Show description for 368f561
  

  2 peopleauthored and

  ericharper

  committed
  368f561

  Copy full SHA for 368f561

• Merge branch 'jbarker/llama3.1' into 'main'

  Show description for c1c19d1
  ericharper

  committed
  c1c19d1

  Copy full SHA for c1c19d1

• ADLR/megatron-lm!2151 - ci: Run experimental UTs on dev image

  ko3n1g

  committed
  1265399

  Copy full SHA for 1265399

• Merge branch 'ko3n1g/ci/uts-on-dev' into 'main'

  Show description for c025cec
  ko3n1g

  committed
  c025cec

  Copy full SHA for c025cec

• ADLR/megatron-lm!1953 - Mcore export to export models to TRTLLM (GPU and CPU version)

  Show description for f0d7120
  

  3 peoplecommitted
  f0d7120

  Copy full SHA for f0d7120

• Merge branch 'final_export' into 'main'

  Show description for 45bf4c1
  shanmugamr1992

  committed
  45bf4c1

  Copy full SHA for 45bf4c1

• ADLR/megatron-lm!2154 - ci: Prune docker cache of mcore-docker-node-jet

  ko3n1g

  committed
  f5171f2

  Copy full SHA for f5171f2

• Merge branch 'ko3n1g/ci/prune-container-cache-mcore-docker-node-jet' into 'main'

  Show description for e38d92a
  ko3n1g

  committed
  e38d92a

  Copy full SHA for e38d92a

• ADLR/megatron-lm!2155 - Resolve release test failure caused by GroupedMLP distributed checkpointing

  

  xxuwenc

  authored and

  ko3n1g

  committed
  c31452c

  Copy full SHA for c31452c

• Merge branch 'xuwenc/release_perf_bugfix' into 'main'

  Show description for d55d61a
  ko3n1g

  committed
  d55d61a

  Copy full SHA for d55d61a

• ADLR/megatron-lm!2156 - tests: Set better name for Wandb logging

  ko3n1g

  committed
  3beefb5

  Copy full SHA for 3beefb5

• Merge branch 'ko3n1g/tests/better-logging-to-wandb' into 'main'

  Show description for 5553fc1
  ko3n1g

  committed
  5553fc1

  Copy full SHA for 5553fc1

Commits on Sep 27, 2024

• ADLR/megatron-lm!1950 - Remove pkg_resources package

  Show description for 0976661
  

  3 peopleauthored and

  ko3n1g

  committed
  0976661

  Copy full SHA for 0976661

• Merge branch 'fix_version_checks' into 'main'

  Show description for 1585be2
  ko3n1g

  committed
  1585be2

  Copy full SHA for 1585be2

• ADLR/megatron-lm!2142 - ci: Onboard CW

  ko3n1g

  committed
  2bad957

  Copy full SHA for 2bad957

• Merge branch 'ko3n1g/ci/onboard-cw' into 'main'

  Show description for 12c2696
  ko3n1g

  committed
  12c2696

  Copy full SHA for 12c2696

Commits on Sep 28, 2024

• ADLR/megatron-lm!2158 - Small changes to export

  Show description for 3428cd9
  

  3 peopleauthored and

  ericharper

  committed
  3428cd9

  Copy full SHA for 3428cd9

• Merge branch 'new_export' into 'main'

  Show description for b3375a0
  ericharper

  committed
  b3375a0

  Copy full SHA for b3375a0

Commits on Sep 30, 2024

• ADLR/megatron-lm!2152 - Fix rope backward compatibility

  

  BoxiangW

  authored and

  jaredcasper

  committed
  5b7374a

  Copy full SHA for 5b7374a

• Merge branch 'boxiangw/mla_backwards_comp' into 'main'

  Show description for 6ad11b0
  jaredcasper

  committed
  6ad11b0

  Copy full SHA for 6ad11b0

Commits on Oct 1, 2024

• ADLR/megatron-lm!2140 - [Bug fix] Don't trace graphs during inference

  

  jiemingz

  authored and

  ericharper

  committed
  ca6d170

  Copy full SHA for ca6d170

• Merge branch 'auto_cudagraph_val_fix' into 'main'

  Show description for dddecd1
  ericharper

  committed
  dddecd1

  Copy full SHA for dddecd1

• ADLR/megatron-lm!2109 - Adding more MR tests for T5 (e.g., transformer_engine, distributed_checkpoint)

  Show description for 5ab659b
  

  2 peopleauthored and

  ko3n1g

  committed
  5ab659b

  Copy full SHA for 5ab659b

• Merge branch 'huvu/t5_dist_checkpoint_mrtests' into 'main'

  Show description for 3efa8c2
  ko3n1g

  committed
  3efa8c2

  Copy full SHA for 3efa8c2

• ADLR/megatron-lm!2164 - ci: Download artifacts

  ko3n1g

  committed
  f07581b

  Copy full SHA for f07581b

• Merge branch 'ko3n1g/ci/artifacts' into 'main'

  Show description for 85cd99b
  ko3n1g

  committed
  85cd99b

  Copy full SHA for 85cd99b

Commits on Oct 2, 2024

• ADLR/megatron-lm!2165 - ci: Bump version

  

  ko3n1g

  authored and

  jaredcasper

  committed
  858694f

  Copy full SHA for 858694f

• Merge branch 'ko3n1g/ci/backwards-tag' into 'main'

  Show description for 065260b
  jaredcasper

  committed
  065260b

  Copy full SHA for 065260b

Commits on Oct 3, 2024

• ADLR/megatron-lm!2153 - Add the interface to set TP communication bootstrap backend

  Show description for f76b465
  

  2 peopleauthored and

  ericharper

  committed
  f76b465

  Copy full SHA for f76b465

• Merge branch 'tp_bootstrap_backend' into 'main'

  Show description for 25f7da2
  ericharper

  committed
  25f7da2

  Copy full SHA for 25f7da2

• ADLR/megatron-lm!2095 - Add support for SigLIP vision encoder to multimodal mcore

  

  Matthieu Le

  authored and

  jaredcasper

  committed
  50042ff

  Copy full SHA for 50042ff

• Merge branch 'convert_siglip_model' into 'main'

  Show description for 4d5f94d
  jaredcasper

  committed
  4d5f94d

  Copy full SHA for 4d5f94d

Commits on Oct 4, 2024

• ADLR/megatron-lm!2175 - adding cu_seqlens_padded support in MCore

  Show description for 2aaf85d
  

  3 peopleauthored and

  ericharper

  committed
  2aaf85d

  Copy full SHA for 2aaf85d

• Merge branch 'add_cu_seqlens_padded_support' into 'main'

  Show description for c02b335
  ericharper

  committed
  c02b335

  Copy full SHA for c02b335

• ADLR/megatron-lm!2181 - Fixing attention mask dimenions to support TE versions > 1.9

  Show description for ee9dba2
  

  2 peopleauthored and

  ericharper

  committed
  ee9dba2

  Copy full SHA for ee9dba2

• Merge branch 'fixattnmask' into 'main'

  Show description for fde8bb1
  ericharper

  committed
  fde8bb1

  Copy full SHA for fde8bb1

• ADLR/megatron-lm!2180 - rotary_scaling fix for llama3.1 and 3.2

  

  yueshen2016

  authored and

  ericharper

  committed
  843a22e

  Copy full SHA for 843a22e

• Merge branch 'yueshen/rotary_scaling_fix_llama3_1' into 'main'

  Show description for b98ec86
  ericharper

  committed
  b98ec86

  Copy full SHA for b98ec86

• ADLR/megatron-lm!2185 - chore: Improve generator for launch scripts

  ko3n1g

  committed
  827d5b6

  Copy full SHA for 827d5b6

• Merge branch 'ko3n1g/ci/fix-launch-script-generator' into 'main'

  Show description for 31fe61a
  ko3n1g

  committed
  31fe61a

  Copy full SHA for 31fe61a

Commits on Oct 5, 2024

• ADLR/megatron-lm!2160 - Adding Inference pipeline for T5

  Show description for e2a1c52
  

  3 peoplecommitted
  e2a1c52

  Copy full SHA for e2a1c52

• Merge branch 'huvu/t5_generate' into 'main'

  Show description for 0acda93
  ericharper

  committed
  0acda93

  Copy full SHA for 0acda93

• ADLR/megatron-lm!2182 - ci: Group runs by model

  ko3n1g

  committed
  2f9ac3c

  Copy full SHA for 2f9ac3c

• Merge branch 'ko3n1g/ci/group-runs' into 'main'

  Show description for edb51fc
  ko3n1g

  committed
  edb51fc

  Copy full SHA for edb51fc

• ADLR/megatron-lm!1862 - Cpu init te

  Show description for cf0d855
  

  4 peopleauthored and

  ko3n1g

  committed
  cf0d855

  Copy full SHA for cf0d855

• Merge branch 'cpu-init-te' into 'main'

  Show description for 0e6bef1
  ko3n1g

  committed
  0e6bef1

  Copy full SHA for 0e6bef1

• ADLR/megatron-lm!2186 - ci: Run script after export

  ko3n1g

  committed
  6939737

  Copy full SHA for 6939737

• Merge branch 'ko3n1g/ci/run-script-after-export' into 'main'

  Show description for 73e7b58
  ko3n1g

  committed
  73e7b58

  Copy full SHA for 73e7b58

Commits on Oct 7, 2024

• ADLR/megatron-lm!2089 - Fix upcycling issues.

  

  RayWang96

  authored and

  ericharper

  committed
  6ca379e

  Copy full SHA for 6ca379e

• Merge branch 'runtime-upcycling' into 'main'

  Show description for ff5cee9
  ericharper

  committed
  ff5cee9

  Copy full SHA for ff5cee9

• ADLR/megatron-lm!2189 - tests: Fix ENV export

  ko3n1g

  committed
  a559ec1

  Copy full SHA for a559ec1

• Merge branch 'ko3n1g/ci/fix-env-export' into 'main'

  Show description for 3f90b98
  ko3n1g

  committed
  3f90b98

  Copy full SHA for 3f90b98

Commits on Oct 9, 2024

• ADLR/megatron-lm!2194 - tests: Fix ENV export

  ko3n1g

  committed
  e108535

  Copy full SHA for e108535

• Merge branch 'ko3n1g/ci/fix-env-export' into 'main'

  Show description for 3f43927
  ko3n1g

  committed
  3f43927

  Copy full SHA for 3f43927

• ADLR/megatron-lm!1790 - GroupedMLP DistOpt Resharding and add UTs to ChainedOptimizer Support for distributed checkpointing

  

  hxbai

  authored and

  ko3n1g

  committed
  fbdc916

  Copy full SHA for fbdc916

• Merge branch 'hongxiaob/moe_dist_ckpt' into 'main'

  Show description for b1218b9
  ko3n1g

  committed
  b1218b9

  Copy full SHA for b1218b9

• ADLR/megatron-lm!2197 - ci: Always upload artifacts

  ko3n1g

  committed
  5776d06

  Copy full SHA for 5776d06

• Merge branch 'ko3n1g/ci/always-artifacts' into 'main'

  Show description for bf74129
  ko3n1g

  committed
  bf74129

  Copy full SHA for bf74129

• ADLR/megatron-lm!2141 - Data parallel inference

  

  trintamaki

  authored and

  jon-barker

  committed
  0e3eaa5

  Copy full SHA for 0e3eaa5

• Merge branch 'trintamaki/data-parallel-inference' into 'main'

  Show description for fcdbf90
  jon-barker

  committed
  fcdbf90

  Copy full SHA for fcdbf90

• ADLR/megatron-lm!2199 - Remove CUDA requirement from cpu test.

  

  Vitaly Kurin

  authored and

  ko3n1g

  committed
  37a2116

  Copy full SHA for 37a2116

• Merge branch 'vitalyk/testfix' into 'main'

  Show description for 228dc20
  ko3n1g

  committed
  228dc20

  Copy full SHA for 228dc20

Commits on Oct 10, 2024

• ADLR/megatron-lm!2096 - Support padding between subsequences of Packed Sequence

  

  parthmannan

  authored and

  ericharper

  committed
  f462160

  Copy full SHA for f462160

• Merge branch 'packed_seq_padded_support' into 'main'

  Show description for 7e90ec0
  ericharper

  committed
  7e90ec0

  Copy full SHA for 7e90ec0

• ADLR/megatron-lm!2206 - Revert "Merge branch 'vitalyk/testfix' into 'main'"

  ko3n1g

  committed
  566d9cd

  Copy full SHA for 566d9cd

• Merge branch 'revert-228dc204' into 'main'

  Show description for b60f5d0
  ko3n1g

  committed
  b60f5d0

  Copy full SHA for b60f5d0

Commits on Oct 11, 2024

• ADLR/megatron-lm!1909 - Standard interface for getting offsets from tokenizers

  

  sancha

  authored and

  ericharper

  committed
  13c39ac

  Copy full SHA for 13c39ac

• Merge branch 'sasatheesh/tokenizer_offsets' into 'main'

  Show description for 47bb8d1
  ericharper

  committed
  47bb8d1

  Copy full SHA for 47bb8d1

• ADLR/megatron-lm!2208 - tests: Use flaky instead of skip marker

  ko3n1g

  committed
  8c018ca

  Copy full SHA for 8c018ca

• Merge branch 'ko3n1g/ci/flaky-marker' into 'main'

  Show description for 772faca
  ko3n1g

  committed
  772faca

  Copy full SHA for 772faca

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

  gurpreet-dhami

  committed
  e8c077c

  Copy full SHA for e8c077c