Fetch from nvidia Megatron-LM by RaymondLi0 · Pull Request #5 · ElementAI/Megatron-LM

RaymondLi0 · 2022-08-03T20:32:25Z

No description provided.

…port full-iteration (FWD-BWD) CUDA graphability. (#4663) Signed-off-by: Cory Ye <cye@nvidia.com>

Signed-off-by: oliver könig <okoenig@nvidia.com>

…5022) Signed-off-by: oliver könig <okoenig@nvidia.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Signed-off-by: oliver könig <okoenig@nvidia.com>

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Signed-off-by: oliver könig <okoenig@nvidia.com>

Signed-off-by: Maanu Grover <maanug@nvidia.com>

…put_parameter (#5034)

…4881) Signed-off-by: Keval Morabia <28916987+kevalmorabia97@users.noreply.github.com> Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

…5040) Signed-off-by: oliver könig <okoenig@nvidia.com>

Signed-off-by: Li Ding <liding@nvidia.com>

Signed-off-by: oliver könig <okoenig@nvidia.com>

#5045) Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Co-authored-by: sraman <sraman@users.noreply.github.com> Co-authored-by: Siddhartha Raman S <sraman@login-lyris01.lyris.clusters.nvidia.com> Co-authored-by: a <a> Co-authored-by: Charlie Truong <chtruong@nvidia.com> Co-authored-by: sraman-rgb <270218152+sraman-rgb@users.noreply.github.com>

Signed-off-by: Yi-Fu Wu <yifu.wu@gmail.com> Co-authored-by: Gerald Shen <geshen@nvidia.com> Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Signed-off-by: nvskills-svc-account <svc-nvskills-signing@nvidia.com> Co-authored-by: nvskills-svc-account <svc-nvskills-signing@nvidia.com>

…l 4768 (#5069) Signed-off-by: oliver könig <okoenig@nvidia.com>

Signed-off-by: Maanu Grover <maanug@nvidia.com>

Signed-off-by: Charlie Truong <chtruong@nvidia.com>

…test (#5512) Signed-off-by: oliver könig <okoenig@nvidia.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Signed-off-by: Deepak Narayanan <dnarayanan@nvidia.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

) Signed-off-by: Antoni-Joan Solergibert <asolergibert@nvidia.com>

Signed-off-by: Lawrence McAfee <lmcafee@nvidia.com>

Signed-off-by: oliver könig <okoenig@nvidia.com>

Signed-off-by: janEbert <janpabloe@nvidia.com>

Signed-off-by: Anil Thomas <anlthms@users.noreply.github.com>

Signed-off-by: Philip Petrakian <ppetrakian@nvidia.com> Co-authored-by: oliver könig <okoenig@nvidia.com>

Signed-off-by: ykarnati <ykarnati@nvidia.com>

#4990) Signed-off-by: Ritesh Patel <ripatel@nvidia.com>

Signed-off-by: Maanu Grover <maanug@nvidia.com>

Signed-off-by: Philip Petrakian <ppetrakian@nvidia.com>

…tures. (#5462) Signed-off-by: Cory Ye <cye@nvidia.com>

#5338) Signed-off-by: ma-jh <3364870135@qq.com>

Signed-off-by: Jennifer Chen <jennifchen@nvidia.com> Signed-off-by: realAsma <akuriparambi@nvidia.com> Co-authored-by: realAsma <akuriparambi@nvidia.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Signed-off-by: Hongbin Liu <hongbinl@nvidia.com> Signed-off-by: hongbinl <hongbinl@nvidia.com> Signed-off-by: Hongbin Liu <lhb8125@users.noreply.github.com> Co-authored-by: Xin Yao <xiny@nvidia.com>

Signed-off-by: ykarnati <ykarnati@nvidia.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Signed-off-by: Jieming Zhang <jiemingz@nvidia.com> Signed-off-by: root <jiemingz@nvidia.com>

Signed-off-by: Jingyue Wu <jingyuew@nvidia.com>

Signed-off-by: Philip Petrakian <ppetrakian@nvidia.com>

Signed-off-by: Ajay Balasa <abalasa@nvidia.com>

ko3n1g force-pushed the main branch from ad5a627 to cd71d4c Compare February 18, 2026 09:18

cspades and others added 29 commits May 27, 2026 15:56

[Megatron-FSDP] Add conditional param.grad dereferencing logic to sup…

67b2f38

…port full-iteration (FWD-BWD) CUDA graphability. (#4663) Signed-off-by: Cory Ye <cye@nvidia.com>

test: restrict iter-time comparison to steady-state window (#5010)

71223d5

Signed-off-by: oliver könig <okoenig@nvidia.com>

fix(test): pin eval-global-batch-size on 15b gb200 release configs (#…

a6d61fb

…5022) Signed-off-by: oliver könig <okoenig@nvidia.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

[fix] Release MTP assertion when EP overlap with PP=1 (#4796)

286445c

fix(test): widen iter-time steady-state window for short tests (#5023)

0cb4034

Signed-off-by: oliver könig <okoenig@nvidia.com>

Perf fix (#4996)

805e24e

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Add dev-feature preservation gate and change schedule (#4773)

457e3f7

chore(test): remove orphan nemotron3_super_release_g200 dir (#5024)

146d171

Signed-off-by: oliver könig <okoenig@nvidia.com>

Ignore Claude worktree directory (#5020)

7be1748

Update copy-pr-bot.yaml [skip ci]

9d23a73

ci: update CI workflow conditions for integration tests (#4658)

3e00820

Add NVSkills CI request workflow (#5033)

0010683

DDP wrap pg size fixes (#5006)

2a4f820

Signed-off-by: Maanu Grover <maanug@nvidia.com>

fix(layer_wise): tag MTP-stage word_embeddings as is_embedding_or_out…

e0afe89

…put_parameter (#5034)

Move LTS dependencies from pyproject.toml to Dockerfile.ci.lts (#4877)

4f2e7ce

Use shared ModelOpt calibration loop on 0.45+ with 0.44 fallback fix (#…

88efdf7

…4881) Signed-off-by: Keval Morabia <28916987+kevalmorabia97@users.noreply.github.com> Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

test(release): skip golden comparison on intermediate resume windows (#…

a5c2d1b

…5040) Signed-off-by: oliver könig <okoenig@nvidia.com>

[mimo] Thread position_ids through MimoModel for multimodal RoPE (#4938)

2ebca1e

Signed-off-by: Li Ding <liding@nvidia.com>

build: Switch DSv3 on H100 to HybridEP (#5039)

e6ff4c3

Signed-off-by: oliver könig <okoenig@nvidia.com>

Fix: Import unwrap_model from megatron.core.utils in modelopt examples (

ba9e0ea

#5045) Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Simple and stable Inference APIs (#4697)

3c39d98

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

ci: Add notification step for MBridge downstream test results (#5028)

f63a46e

Delete output tensor early (#4742)

f1b5516

Skip gradient updates when grad norm exceeds threshold (#3460)

1801316

Signed-off-by: Yi-Fu Wu <yifu.wu@gmail.com> Co-authored-by: Gerald Shen <geshen@nvidia.com> Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Add 9 user skills (#5066)

6e0d14a

Signed-off-by: nvskills-svc-account <svc-nvskills-signing@nvidia.com> Co-authored-by: nvskills-svc-account <svc-nvskills-signing@nvidia.com>

test(nemotron): align nemotron3 super GB200 goldens with exit-interva…

54dc530

…l 4768 (#5069) Signed-off-by: oliver könig <okoenig@nvidia.com>

chore: Update transformer-engine dependency to version 2.16.0 (#4992)

6e091e1

Update energon version requirement (#4572)

52d1d68

Signed-off-by: Maanu Grover <maanug@nvidia.com>

chtruong814 and others added 30 commits June 27, 2026 07:51

ci: Use GB300 for Github CI tests (#5520)

ed59a0a

Signed-off-by: Charlie Truong <chtruong@nvidia.com>

ci: pin HF_HUB_CACHE to bind-mounted cache for gpt-oss-20b inference …

0ff7226

…test (#5512) Signed-off-by: oliver könig <okoenig@nvidia.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Add inter-document attention masking to GPTDataset (#5298)

f88b85f

Signed-off-by: Deepak Narayanan <dnarayanan@nvidia.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Update copy-pr-bot.yaml [skip ci]

25f6a09

[CI] Fix gpt_dynamic_inference_tp2_pp2_ep2_gptoss_20b_swa tests (#5527

522a9dd

) Signed-off-by: Antoni-Joan Solergibert <asolergibert@nvidia.com>

Implement async scheduling for dynamic inference (#5453)

3c327f3

Signed-off-by: Lawrence McAfee <lmcafee@nvidia.com>

build: bump transformer-engine to release_v2.16.post (#5517)

b9f7fb6

Signed-off-by: oliver könig <okoenig@nvidia.com>

Fix isort target Python version (#5567)

819912c

Signed-off-by: janEbert <janpabloe@nvidia.com>

Fix PR template typo (#5566)

7b6eb02

Signed-off-by: janEbert <janpabloe@nvidia.com>

Deduplicate tensor-splitting utility (#5545)

872442a

Signed-off-by: Anil Thomas <anlthms@users.noreply.github.com>

Add CI duties to oncall (#5510)

99b56a7

Signed-off-by: Philip Petrakian <ppetrakian@nvidia.com> Co-authored-by: oliver könig <okoenig@nvidia.com>

Thread dp_cp/expt_dp process groups through checkpoint load path (#5579)

817c1d5

Signed-off-by: ykarnati <ykarnati@nvidia.com>

Fix TEGroupedMLP pre-backward unshard in fine-grained FSDP hooks for … (

f285ea5

#4990) Signed-off-by: Ritesh Patel <ripatel@nvidia.com>

[training migration] Finish ModelBuilder integration (#5516)

223e244

Signed-off-by: Maanu Grover <maanug@nvidia.com>

Update copy-pr-bot.yaml [skip ci]

4849143

Use NVIDIA inference credentials for Claude actions (#5589)

a3d761b

Signed-off-by: Philip Petrakian <ppetrakian@nvidia.com>

Update PR instructions (#5592)

36d11ce

Signed-off-by: Philip Petrakian <ppetrakian@nvidia.com>

Update mcore skill owners (#5586)

f972da7

Signed-off-by: Philip Petrakian <ppetrakian@nvidia.com>

chore: rotate oncall schedule

8b3d8a5

Add /claude fix workflow for on-demand PR fixes (#4862)

3c08255

Signed-off-by: Philip Petrakian <ppetrakian@nvidia.com>

Update copy-pr-bot.yaml [skip ci]

4c4a8ee

[Megatron-FSDP] MaxPoolAllocator for double-buffering hybrid architec…

adfe9e1

…tures. (#5462) Signed-off-by: Cory Ye <cye@nvidia.com>

fix(tensor_parallel): _reduce returns unreduced tensor for non-contig… (

60f338a

#5338) Signed-off-by: ma-jh <3364870135@qq.com>

Add Auto Quantize in ModelOpt quantize example (#4821)

2b551c6

Signed-off-by: Jennifer Chen <jennifchen@nvidia.com> Signed-off-by: realAsma <akuriparambi@nvidia.com> Co-authored-by: realAsma <akuriparambi@nvidia.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

[Main][feat] Support CUDA Graph capture offloading modules (#3697)

69c4868

Signed-off-by: Hongbin Liu <hongbinl@nvidia.com> Signed-off-by: hongbinl <hongbinl@nvidia.com> Signed-off-by: Hongbin Liu <lhb8125@users.noreply.github.com> Co-authored-by: Xin Yao <xiny@nvidia.com>

E2E heterogenous non colocated MiMo training (#5602)

0522099

Signed-off-by: ykarnati <ykarnati@nvidia.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Optimize memory usage of partial CUDA graphs (#5451)

5e4fe9b

Signed-off-by: Jieming Zhang <jiemingz@nvidia.com> Signed-off-by: root <jiemingz@nvidia.com>

Document stacked dependent PR handling in split PR skill (#5496)

ee623a9

Signed-off-by: Jingyue Wu <jingyuew@nvidia.com>

Fix Claude reaction permissions (#5613)

060371c

Signed-off-by: Philip Petrakian <ppetrakian@nvidia.com>

Fix smoke BERT/T5 test failures (#5629)

6608a05

Signed-off-by: Ajay Balasa <abalasa@nvidia.com>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Fetch from nvidia Megatron-LM#5

Fetch from nvidia Megatron-LM#5
RaymondLi0 wants to merge 7611 commits into
ElementAI:load-iterfrom
NVIDIA:main

RaymondLi0 commented Aug 3, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

19 participants

Uh oh!

Conversation

RaymondLi0 commented Aug 3, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

19 participants