Skip to content

Pull from Main #956

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
YianZhang wants to merge 35 commits into from
Closed

Pull from Main #956

YianZhang wants to merge 35 commits into from

Conversation

@YianZhang
Copy link

@YianZhang YianZhang commented Aug 20, 2025

No description provided.

ZhiyuLi-Nvidia and others added 30 commits August 4, 2025 19:15
@ZhiyuLi-Nvidia
fix: assert no duplicate starting bos (#835)
84abe2c 
Signed-off-by: Zhiyu Li <[email protected]>
@ahmadki
fix: Added sequence packing keys to SFT and GRPO recipes (#805)
fa21d3c 
Signed-off-by: Ahmad Kiswani <[email protected]>
@ahmadki
fix: OOM with some GRPO configs (#709)
5604024 
Signed-off-by: Ahmad Kiswani <[email protected]>
@yuki-97
chore: upgrade vllm to v0.10.0 (#766)
9389dfb 
Signed-off-by: Yuki Huang <[email protected]>
@abdalgader-a
docs: fix checkpointing command for megatron->hf export (#823)
c249efc 
Signed-off-by: abdalgader-a <[email protected]>
@ZhiyuLi-Nvidia
feat: add data shuffle and random seed option (#334)
c784dd9 
Signed-off-by: Zhiyu Li <[email protected]>
Signed-off-by: Zhiyu Li <[email protected]>
@wedu-nvidia @terrykong
feat: save checkpoint before timeout to avoid 4-hour runtime limit (#734
b74c5d0 
)

Signed-off-by: Wei Du <[email protected]>
Signed-off-by: Terry Kong <[email protected]>
Co-authored-by: Terry Kong <[email protected]>
@ybgao-nvidia
feat: track policy training compute throughput (#632)
b6269f7 
Signed-off-by: Yubo Gao <[email protected]>
@ashors1
fix: fix grpo + mcore checkpointing without validation (#844)
9af0a52 
Signed-off-by: ashors1 <[email protected]>
@terrykong
feat: dockerfile can build hermetically or from build context (#799)
03472a0 
Signed-off-by: Terry Kong <[email protected]>
@terrykong
chore: 0.3.0 -> 0.4.0rc0 (#840)
0557402 
Signed-off-by: Terry Kong <[email protected]>
@parthchadha
fix: force use of eager (disabled cuda graphs) due to convergence iss…
233cc07 
…ues (#857)

Signed-off-by: Parth Chadha <[email protected]>
@ffrujeri
fix: Fix error message in VllmGenerationWorker. (#633)
0988a7d 
Signed-off-by: Felipe Vieira Frujeri <[email protected]>
@ashors1
feat: support DTensor CP in DPO and SFT (#798)
5910abb 
Signed-off-by: ashors1 <[email protected]>
@yuki-97
feat: support non-colocated in mcore (#613)
b8a89a9 
Signed-off-by: Yuki Huang <[email protected]>
@yuki-97
chore: remove old fsdp1 unit test (#871)
88a399e 
Signed-off-by: Yuki Huang <[email protected]>
@yuki-97
fix: fix non-colocated with cpu_offload enabled (#861)
bbbb3d6 
Signed-off-by: Yuki Huang <[email protected]>
@wangshangsam
docs: Link uv's installation instructions to uv's website (#837)
e924d33 
Signed-off-by: Shang Wang <[email protected]>
@ashors1 @abdalgader-a
feat: qwen3 export to HF (#873)
d73c942 
Signed-off-by: Abdalgader Abubaker <[email protected]>
Signed-off-by: Anna Shors <[email protected]>
Co-authored-by: Abdalgader Abubaker <[email protected]>
@terrykong
test: add deepscaler tests + pipe-clean configs + fix eval for deepsc…
d45ff3f 
…aler (#866)

Signed-off-by: Terry Kong <[email protected]>
@RayenTian
fix: remove tie weight check (#700)
fecf71e 
Signed-off-by: ruit <[email protected]>
@soodoshll
fix: OOM in deepscaler1.5b with sequence length = 16/24k (#875)
2b87def 
Signed-off-by: Qidong Su <[email protected]>
@guyueh1
feat: Fix and enhances for Nsight system profiling (#865)
8fd8c96 
Signed-off-by: Guyue Huang <[email protected]>
@terrykong
test: lower step count on gemma nightly test to finish within 4 hours (…
18b9e2c 
…#880)

Signed-off-by: Terry Kong <[email protected]>
@gshennvm @terrykong
feat: add nemotron5 sharding (#481)
223bfa8 
Signed-off-by: Terry Kong <[email protected]>
Co-authored-by: Terry Kong <[email protected]>
@terrykong
feat: add diagnostic script for problematic embeddings (#896)
e1f56c4 
Signed-off-by: Terry Kong <[email protected]>
@RayenTian
feat: Add TP to embed_tokens and lm_head for Gemma models (#879)
9f7825e 
Signed-off-by: ruit <[email protected]>
@yuki-97
refactor: split sync/async vllm worker ([1/2] of refactor vllm worker) (
83c6bfc 
#900)

Signed-off-by: Yuki Huang <[email protected]>
@pjin-nvidia
feat: chunked logprob calculation with deferred fp32 cast to help wit…
df31c1b 
…h OOM (#918)

Signed-off-by: Peter Jin <[email protected]>
@chtruong814
build: Add Dockerfile that uses NGC pytorch image (#897)
70b9666 
Signed-off-by: Charlie Truong <[email protected]>
chtruong814 and others added 5 commits August 18, 2025 16:22
@chtruong814
build: Fix pytorch image ref in Dockerfile.ngc_pytorch (#936)
b721703 
Signed-off-by: Charlie Truong <[email protected]>
@terrykong
test: enable 8k/16k/24k deepscaler nightly tests (#934)
d149a62 
Signed-off-by: Terry Kong <[email protected]>
@rohitrango @yfw
@terrykong @chtruong814
feat: GRPO + SFT Dtensor support for multimodal training (#712)
eb50202 
Signed-off-by: Yi-Fu Wu <[email protected]>
Signed-off-by: rohitrango <[email protected]>
Signed-off-by: Terry Kong <[email protected]>
Signed-off-by: Charlie Truong <[email protected]>
Co-authored-by: Yi-Fu Wu <[email protected]>
Co-authored-by: Terry Kong <[email protected]>
Co-authored-by: Charlie Truong <[email protected]>
@mrm-196
feat: Add recipe to reproduce Tulu-3 DPO model (#804)
bde1a68 
Signed-off-by: root <[email protected]>
Signed-off-by: root <[email protected]>
Signed-off-by: root <[email protected]>
Signed-off-by: root <[email protected]>
Co-authored-by: root <[email protected]>
Co-authored-by: root <[email protected]>
Co-authored-by: root <[email protected]>
Co-authored-by: root <[email protected]>
@chtruong814
ci: Fix docker build context (#942)
f85a4f6 
Signed-off-by: Charlie Truong <[email protected]>
@github-actions github-actions bot added documentation Improvements or additions to documentation CI Relating to CI labels Aug 20, 2025
@YianZhang YianZhang closed this Aug 20, 2025
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

CI Relating to CI documentation Improvements or additions to documentation

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.