Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
307 Open 1,657 Closed
307 Open 1,657 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

Multimodal: add load weights only
#3452 opened Feb 17, 2026 by faradawn Loading…
6 tasks
1
1
Multimodal: fix argument checking
#3449 opened Feb 17, 2026 by faradawn Loading…
6 tasks
Core 0.16
2
Multimodal: Limit transformer version in Dockerfile
#3448 opened Feb 17, 2026 by faradawn Loading…
6 tasks
1
Add mxfp8 quantization for inference linear layers
#3447 opened Feb 17, 2026 by santhnm2 • Draft
6 tasks
1
Log RL metrics per environment
#3446 opened Feb 16, 2026 by yobibyte Loading…
2
Fix slowdown in inference flask server
#3445 opened Feb 16, 2026 by tdene Loading…
6 tasks
Core 0.16
@santhnm2
ci(fix): Weekly GPT tests
#3443 opened Feb 16, 2026 by ko3n1g Loading…
6 tasks
Core 0.16
Adding ssm kernels to support packed sequences with initial states
#3442 opened Feb 16, 2026 by oyilmaz-nvidia • Draft
5 tasks
1
update HF tokenizer defaults
#3440 opened Feb 16, 2026 by dimapihtar Loading…
6 tasks
Core 0.16
8
fix traceback when interrupting run
#3439 opened Feb 16, 2026 by dimapihtar • Draft
6 tasks
1
ci: Update release-docs workflow to use FW-CI-templates v0.72.0 core_r0.16.0 Cherry-pick label for core_r0.16.0 release branch
#3438 opened Feb 15, 2026 by chtruong814 Loading…
6 tasks
Core 0.16
Clean up logging inside inference flask server
#3437 opened Feb 15, 2026 by tdene Loading…
6 tasks
Core 0.16
1
Create a Protocol for the MLP layer of TransformerLayer community-request complexity: medium Expert Review Apply this label to indicate that your PR is ready for expert review.
#3435 opened Feb 15, 2026 by nschank Loading…
2 of 6 tasks
Core 0.16
2
Use Protocols to type-check linear_proj submodules of Attention community-request complexity: medium Expert Review Apply this label to indicate that your PR is ready for expert review.
#3434 opened Feb 15, 2026 by nschank Loading…
2 of 6 tasks
Core 0.16
2
[main] Refactor and Improve MoE Logginginit commit
#3431 opened Feb 14, 2026 by yanring • Draft
6 tasks
1
Feature: PyTorch Native Implementation of Manifold Hyper Connection
#3430 opened Feb 14, 2026 by jingqiny-99 • Draft
6 tasks
1
Enable CUDA graph for ADAM optimizer
#3429 opened Feb 14, 2026 by vasunvidia Loading…
6 tasks
1
Fix LLaVA packed sequence cu_seqlens after image token expansion
#3427 opened Feb 14, 2026 by aroshanghias-nvd Loading…
2 of 4 tasks
1
Migrate MoeLayer submodules from ModuleSpec to Protocols community-request complexity: medium Expert Review Apply this label to indicate that your PR is ready for expert review.
#3426 opened Feb 14, 2026 by nschank Loading…
2 of 6 tasks
Core 0.16
3
Ensure type-checker understands use of Submodules in unit tests community-request needs-follow-up Issue needs follow-up
#3425 opened Feb 13, 2026 by nschank Loading…
2 of 6 tasks
1
Add single-process checkpoint save to avoid forked multiprocessing
#3424 opened Feb 13, 2026 by sbak5 Loading…
6 tasks
1
Fix --tokenizer-hf-include-special-tokens Expert Review Apply this label to indicate that your PR is ready for expert review. Run functional tests Run tests
#3422 opened Feb 13, 2026 by jon-barker Loading…
2 of 6 tasks
Core 0.16
@jon-barker
11
Use copy_signature to preserve typing of pass-through methods community-request complexity: low Expert Review Apply this label to indicate that your PR is ready for expert review.
#3419 opened Feb 13, 2026 by nschank Loading…
2 of 6 tasks
Core 0.16
2
Improve Multimodal README: add pointer to Qwen3 VL example
#3418 opened Feb 13, 2026 by faradawn Loading…
1 of 6 tasks
1
Bug fix: add missing packages to Multimodal Dockerfile
#3417 opened Feb 13, 2026 by faradawn Loading…
1 of 6 tasks
1
1
ProTip! What’s not been updated in a month: updated:<2026-01-17.