Skip to content

Fetch from nvidia Megatron-LM#5

Open
RaymondLi0 wants to merge 6813 commits intoElementAI:load-iterfrom
NVIDIA:main
Open

Fetch from nvidia Megatron-LM#5
RaymondLi0 wants to merge 6813 commits intoElementAI:load-iterfrom
NVIDIA:main

Conversation

@RaymondLi0
Copy link

@RaymondLi0 RaymondLi0 commented Aug 3, 2022

No description provided.

ahmadki and others added 30 commits February 9, 2026 06:47
@ahmadki @yaox12
General README and pyproject fixes (#2907)
3069591 
Co-authored-by: Xin Yao <xiny@nvidia.com>
@ko3n1g
chore: More aggressive checkpointing (#3315)
3bb539e 
Signed-off-by: oliver könig <okoenig@nvidia.com>
@ko3n1g
ci: Pin down setuptools to lt 82 (#3313)
c072f89 
Signed-off-by: oliver könig <okoenig@nvidia.com>
@ko3n1g
fix: T5 dataset (#3307)
9ddbce3 
Signed-off-by: oliver könig <okoenig@nvidia.com>
@ko3n1g
fix: numpy overflow (#3306)
f14d161 
Signed-off-by: oliver könig <okoenig@nvidia.com>
@chtruong814
ci: Revert "ci: Add secrets detector (#3180)" (#3330)
8d79987 
Signed-off-by: Charlie Truong <chtruong@nvidia.com>
@ko3n1g
ci: Add more tests, run on merge-queue (#3317)
55c3e63 
Signed-off-by: oliver könig <okoenig@nvidia.com>
@chtruong814
ci: Remove merge-gate environment check (#3331)
ba76934 
Signed-off-by: Charlie Truong <chtruong@nvidia.com>
@kwyss-nvidia @yaox12 @ericharper
Use FP4 context for mamba (#2604)
ab5e277 
Co-authored-by: Xin Yao <xiny@nvidia.com>
Co-authored-by: Eric Harper <eharper@nvidia.com>
@chtruong814
ci: Ensure we run all functional tests in merge group (#3332)
fc557ec 
Signed-off-by: Charlie Truong <chtruong@nvidia.com>
@nschank
Replace ModuleSpec with Protocols for inputs to MLP (#3084)
55198ba
@chtruong814
ci: Fix merge queue functional tests (#3337)
5eb20b8 
Signed-off-by: Charlie Truong <chtruong@nvidia.com>
@ko3n1g
ci: skip queue in merge-gate (#3343)
367f0b8 
Signed-off-by: oliver könig <okoenig@nvidia.com>
@ko3n1g
ci: Timeout for functional tests (#3346)
3fb6006 
Signed-off-by: oliver könig <okoenig@nvidia.com>
@dimapihtar
update checkpointing documentation (#3347)
76cf11e 
Signed-off-by: dimapihtar <dpihtar@gmail.com>
@tdene
Update golden values to reflect improvements (#3350)
836d473
@sancha @BestJuly
BUGFIX: gpt vs hybrid model mtp naming mismatch (#3334)
2451508 
Co-authored-by: Li Tao <lit@nvidia.com>
@tdene
Disable flaky test (#3354)
8da949e
@jon-barker
re-enable gpt grpo tests (#3348)
bb97791
@asolergi-nv
Fix SFT Pipeline when TP>1 (#3268)
4bce841
@AAnoosheh
Fixes for KD mode (#3342)
f5238ba 
Signed-off-by: Asha Anoosheh <aanoosheh@nvidia.com>
@github-actions
chore: rotate oncall schedule
c1169ea
@ko3n1g
chore: Update codeowners file (#3365)
4f025a1 
Signed-off-by: oliver könig <okoenig@nvidia.com>
@sidsingh-nvidia @ko3n1g
Siddharth/fix inference functional tests (#3357)
66ec17e 
Co-authored-by: oliver könig <okoenig@nvidia.com>
@janEbert @ko3n1g
Switch oncall (#3360)
6a9da99 
Co-authored-by: oliver könig <okoenig@nvidia.com>
@AAnoosheh
Add missing RMSNorm to llama train script (#3314)
b6e883b 
Signed-off-by: Asha Anoosheh <aanoosheh@nvidia.com>
@tdene
Fix inference for MTP models (#3297)
cd14090
@yobibyte
Add a logprobs test with real gpt model. (#2870)
6f5de16
@tdene
Add simple GRPO functional test (#3323)
b5d50cb
@ko3n1g
ci: Concurrency control for merge-queue (#3353)
1c245c7 
Signed-off-by: oliver könig <okoenig@nvidia.com>
CreeperLKF and others added 30 commits March 6, 2026 18:48
@CreeperLKF @Phlip79
fix: temperature validation error message 1000.0 -> 100.0 (#2688)
26f9444 
Co-authored-by: Philip Petrakian <ppetrakian@nvidia.com>
@mathemakitten
RL: Hybrid MoE training cudagraphs and fix training <-> inference tra…
e19fbe2 
…nsition (#3373)
@santhnm2
Fix dynamic inference and GRPO functional tests (#3740)
c1e675f 
Signed-off-by: Keshav Santhanam <ksanthanam@nvidia.com>
@janEbert
Swap oncall (#3585)
0cfa420
@github-actions
Update copy-pr-bot.yaml [skip ci]
932b767
@leisuzz @Phlip79
[bugfix] fix the bug that loss: 0 will not be printed (#1555)
b09ee64 
Co-authored-by: Philip Petrakian <ppetrakian@nvidia.com>
@CarlosGomes98 @ericharper
@claude @Phlip79
Fused dLN + add in backwards pass (#3384)
8318b80 
Co-authored-by: Eric Harper <complex451@gmail.com>
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
Co-authored-by: Philip Petrakian <ppetrakian@nvidia.com>
@github-actions
chore(beep boop 🤖): Bump (main) (2026-03-09)
597721a
@Phlip79
Claude: run actions on target branch (#3745)
116a7fa
@dimapihtar
revert of #2658 (#3736)
56158bb 
Signed-off-by: dimapihtar <dpihtar@gmail.com>
@ilml @ericharper
Update README Quick Start (#3596)
452fc11 
Co-authored-by: Eric Harper <eharper@nvidia.com>
@mathemakitten
Re-enable tests which were failing on #3373 (#3757)
d904a68
@Phlip79
Check reviews properly (#3756)
94ff0dc
@github-actions
Update copy-pr-bot.yaml [skip ci]
ce66b22
@mehraakash
Add CP + Sequence Packing support for Mimo (#2135)
0e19bf1
@wdykas @mathemakitten
@ahmadki @cursoragent
MXFP8 refit (#3742)
fca1679 
Signed-off-by: Ahmad Kiswani <kiswani.ahmad@gmail.com>
Co-authored-by: Helen Ngo <helenn@nvidia.com>
Co-authored-by: Shanmugam Ramasamy <shanmugamr@cw-dfw-cs-001-login-01.cm.cluster>
Co-authored-by: Ahmad Kiswani <kiswani.ahmad@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: William Dykas <wdykas@oci-hsg-cs-001-vscode-03.cm.cluster>
Co-authored-by: root <root@nvl72136-T10.cm.cluster>
@Phlip79
Claude: update token usage (#3760)
397772e
@sancha
Handle Tool Call Argument Parsing (#3662)
3eea580
@jon-barker
RL support for nanov3 sft checkpoint (#3741)
e970199 
Co-authored-by: Jon Barker <jbarker@oci-hsg-cs-001-vscode-01.cm.cluster>
@yeyu-nvidia
add mix_hidden_states option in conversion (#3655)
ba497c9
@ko3n1g
ci: Optimize release-configs for GB200 (#3541)
22c69fa 
Signed-off-by: oliver könig <okoenig@nvidia.com>
@github-actions
Update copy-pr-bot.yaml [skip ci]
0f47a1a
@kunlunl
Add absorbed-mla (#3198)
204e7d5
@Victarry @claude
feat(checkpoint): zero-copy storage sharing in CheckpointWithoutOutput (
f544034 
#3649)

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@cjld @yaox12
Fuse MLA DOWN projection GEMMs (#3039)
8fd390d 
Signed-off-by: Xin Yao <xiny@nvidia.com>
Co-authored-by: Xin Yao <xiny@nvidia.com>
@Victarry @claude
fix: skip FSDP DTensor boundary validation under fake process group (#…
7d52694 
…3686)

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@yuzhongw-nvidia
[main] fix(moe): fix the bug where gate was not sliced when kv_head <…
16a8cdb 
… tp_size. (#3575)

Co-authored-by: xiaotaoliu <xiaotaoliu@tencent.com>
@rapatel @lhb8125
fix(offload): reset activation offload manager after eval as well as … (
b8e23d5 
#3739)

Co-authored-by: Hongbin Liu <lhb8125@users.noreply.github.com>
@github-actions
chore: rotate oncall schedule
07e512a
@yobibyte
Improve error logging when invalid number of tokens is requested. (#3680
e20b89a 
)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.