Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
331 Open 1,863 Closed
331 Open 1,863 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Support mxfp8 proj gemm weight quant caching
#4489 opened Apr 28, 2026 by gdengk Contributor Loading…
5 tasks
1
cp: add permute fusion into hybrid ep (4089) into core_r0.17.0 cherry-pick Run CICD
#4488 opened Apr 28, 2026 by ko3n1g Contributor Loading… Core 0.16
2
test: skip mfsdp_fully_shard cases when world_size < mesh size complexity: low Run tests
#4487 opened Apr 27, 2026 by wujingyue Contributor Loading…
1 of 2 tasks
Core 0.16
2
Add preliminary Muon+M-FSDP support
#4486 opened Apr 27, 2026 by janEbert Contributor • Draft
1
Standardize misc graph interface complexity: medium
#4485 opened Apr 27, 2026 by tdene Contributor Loading…
5 tasks
Core 0.16
2
Add mHC transformer reference implementation Run functional tests
#4483 opened Apr 27, 2026 by Connor-XY Loading… Core 0.16
83
Checkpoint conversion between GPT_model and Hybrid_model
#4482 opened Apr 27, 2026 by guihong-nv Contributor • Draft
1 of 5 tasks
1
[dev] [DeepSeek-v4] Part 2: Hash MoE, SwiGLU clamp, and new mHC contract dev branch Dev branch related issues and development
#4481 opened Apr 27, 2026 by hxbai Contributor • Draft
5 tasks
@hxbai
1
chore(beep boop 🤖): Bump uv.lock (core_r0.17.0) (2026-04-27)
#4478 opened Apr 27, 2026 by svcnvidia-nemo-ci Loading…
1 task
Core 0.16
2
Start draft PR for MLA support to Muon optimizer community-request
#4477 opened Apr 26, 2026 by Prachi-kushwaha • Draft
5 tasks
2
Start draft PR for get_tensor_device fix community-request
#4476 opened Apr 26, 2026 by Prachi-kushwaha Loading…
2
[Draft] Muon CPU offload community-request
#4475 opened Apr 26, 2026 by pengdurice • Draft
5 tasks
1
feat(attention): Add attention_per_head_gate and rotary_base_per_laye…
#4473 opened Apr 26, 2026 by shifangx Contributor Loading…
5 tasks
8
docs(pipeline_parallel): clarify seq_length behavior with variable_seq_lengths under PP community-request
#4471 opened Apr 25, 2026 by edenfunf • Draft
2 of 3 tasks
1
1
docs(moe): correct moe_router_topk_scaling_factor docstring community-request complexity: low waiting-on-customer Waiting on the original author to respond
#4470 opened Apr 25, 2026 by edenfunf Loading… Core 0.16
1
3
Add mHC support for HybridModel on dsv4 complexity: high Run functional tests
#4469 opened Apr 24, 2026 by Connor-XY Loading… Core 0.16
61
[codex] Fix Mamba conv params under fine-grained FSDP gather complexity: low
#4467 opened Apr 24, 2026 by ilml Contributor Loading…
6
Inference: reduce EP consensus frequency when work is in flight complexity: low
#4464 opened Apr 24, 2026 by sidsingh-nvidia Contributor Loading…
5 tasks
Core 0.16
2
Add Hybrid Transformer block fusion complexity: high
#4463 opened Apr 24, 2026 by janEbert Contributor Loading… Core 0.16
3
ci: Fix event name reference in CI workflow condition for merge group Approved All necessary approvals have been made complexity: low
#4462 opened Apr 24, 2026 by balasaajay Contributor Loading…
5 tasks
Core 0.16
5
mamba: shift silu(z) gate from RMSNormGated into selective_state_update
#4461 opened Apr 24, 2026 by wdykas Contributor • Draft
5 tasks
6
mamba: avoid redundant HBM reloads in causal_conv1d_update shift loop
#4460 opened Apr 24, 2026 by wdykas Contributor • Draft
5 tasks
Core 0.16
6
Fused add rmsnorm
#4459 opened Apr 24, 2026 by wdykas Contributor • Draft
5 tasks
Core 0.16
3
[dev] [DeepSeek-v4] Part 1: Hybrid Attention with CSA and HCA dev branch Dev branch related issues and development
#4458 opened Apr 24, 2026 by hxbai Contributor • Draft
5 tasks
@hxbai
4
[PP] Add initial overlap_p2p_comm support for non-interleaved steady-state 1F1B community-request
#4456 opened Apr 24, 2026 by cky-dev • Draft
1
3
ProTip! no:milestone will show everything without a milestone.