Skip to content

Pull requests: NVIDIA-NeMo/Megatron-Bridge

New pull request New
211 Open 5,053 Closed
211 Open 5,053 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

fix(ci): disable optional NCCL EP in wheel packaging
#6165 opened Sep 20, 2026 by cuichenx Contributor • Draft
2
fix(deepseek): preserve main and dev conversion compatibility
#6164 opened Sep 20, 2026 by cuichenx Contributor • Draft
2
feat(deepseek-v4): let inference.sh launch across nodes area:model Model implementations and HF bridge logic community-request feature New capabilities, enhancements, or enablement work needs-more-tests Requires additional L0 and L1 test coverage before merge needs-review PR is ready for code review and waiting on a reviewer
#6163 opened Sep 20, 2026 by linmuchuiyang Loading…
2 of 4 tasks
1
fix(models): default the HF config file lock to HF_HOME instead of ~/.cache/huggingface area:model Model implementations and HF bridge logic bug Something isn't working community-request needs-review PR is ready for code review and waiting on a reviewer
#6162 opened Sep 20, 2026 by linmuchuiyang Loading…
3 of 4 tasks
1
[build] chore: bump MCore main to d476c21 area:build Dependencies, packaging, images, and environment setup area:ckpt Checkpoint conversion, loading, export, and save paths blocked Work cannot move forward until an external dependency is cleared ci CI, automation, test queue, or workflow infrastructure work full-test-suite needs-more-tests Requires additional L0 and L1 test coverage before merge
#6160 opened Sep 19, 2026 by cuichenx Contributor Loading…
7
docs: add 0.7.0 to version picker area:misc Cross-cutting utilities, logging, helpers, and other changes docs Documentation-only updates or documentation debt docs-only With great power comes great responsibility. needs-review PR is ready for code review and waiting on a reviewer
#6159 opened Sep 19, 2026 by balasaajay Contributor Loading…
[Draft] Migrate GLM from GPTModel to HybridModel
#6158 opened Sep 19, 2026 by xuantengh • Draft
5 tasks
1
fix(training): pass explicit groups through MLM compatibility providers
#6151 opened Sep 18, 2026 by Connor-XY • Draft
1
fix(moe): require explicit dispatcher backend selection area:training Training loop, callbacks, and runtime integration breaking-change Public behavior or API compatibility changes bug Something isn't working full-test-suite needs-more-tests Requires additional L0 and L1 test coverage before merge needs-review PR is ready for code review and waiting on a reviewer
#6150 opened Sep 18, 2026 by cuichenx Contributor Loading…
1
1
feat(training): log DSA indexer loss via MCore helper area:training Training loop, callbacks, and runtime integration community-request feature New capabilities, enhancements, or enablement work needs-more-tests Requires additional L0 and L1 test coverage before merge needs-review PR is ready for code review and waiting on a reviewer
#6145 opened Sep 18, 2026 by lkhphuc Loading…
2 of 5 tasks
1
Delay Megamoe wgrads community-request
#6143 opened Sep 18, 2026 by vthumbe1503 • Draft
5 tasks
1
fix(ckpt): thread revision through SafeTensorsStateSource area:ckpt Checkpoint conversion, loading, export, and save paths bug Something isn't working community-request needs-review PR is ready for code review and waiting on a reviewer
#6142 opened Sep 17, 2026 by Lingikaushikreddy Contributor Loading…
4 tasks done
1
1
feat(data): support custom Energon task encoder configs area:data Dataset builders, preprocessing, and samplers feature New capabilities, enhancements, or enablement work needs-more-tests Requires additional L0 and L1 test coverage before merge needs-review PR is ready for code review and waiting on a reviewer
#6141 opened Sep 17, 2026 by cuichenx Contributor Loading…
3 of 4 tasks
1
2
Fix(qwen3vl): finalize TE precision after decoder replacement area:model Model implementations and HF bridge logic bug Something isn't working needs-more-tests Requires additional L0 and L1 test coverage before merge needs-review PR is ready for code review and waiting on a reviewer
#6140 opened Sep 17, 2026 by vasunvidia Contributor Loading…
5 tasks
2
reuse run_config from MLM
#6139 opened Sep 17, 2026 by dimapihtar Contributor • Draft
5 tasks
1
feat(glm): GLM-5.3-Flash (glm5_next) bridge, provider and block spec community-request
#6137 opened Sep 17, 2026 by erictang000 Contributor • Draft
5 tasks done
1
feat(training): apply shared MCore determinism startup policy area:training Training loop, callbacks, and runtime integration feature New capabilities, enhancements, or enablement work
#6134 opened Sep 17, 2026 by Connor-XY • Draft
3 of 5 tasks
1
feat(ckpt): add nemotron super vl text-only conversion area:ckpt Checkpoint conversion, loading, export, and save paths feature New capabilities, enhancements, or enablement work
#6130 opened Sep 17, 2026 by cuichenx Contributor • 2/2 • Draft
3 of 5 tasks
1
feat(training): online sequence packing and dynamic context parallel with a Qwen3.5-35B-A3B recipe area:training Training loop, callbacks, and runtime integration blocked Work cannot move forward until an external dependency is cleared feature New capabilities, enhancements, or enablement work full-test-suite high-complexity Harder to merge: prone to conflicts and needs additional test coverage needs-more-tests Requires additional L0 and L1 test coverage before merge
#6127 opened Sep 16, 2026 by ilml Contributor Loading…
6 tasks done
2
Expose immutable rank-local LoRA export for native transport community-request
#6126 opened Sep 16, 2026 by hershg Contributor • Draft
1
feat(data): ragged on-disk store for packed SFT materialization area:data Dataset builders, preprocessing, and samplers community-request feature New capabilities, enhancements, or enablement work needs-more-tests Requires additional L0 and L1 test coverage before merge needs-review PR is ready for code review and waiting on a reviewer
#6125 opened Sep 16, 2026 by kenkenpa2126 Contributor Loading…
4 of 5 tasks
1
feat(qwen): add Qwen4-Exp (Qwen3.8-Flash-Next) area:model Model implementations and HF bridge logic blocked Work cannot move forward until an external dependency is cleared community-request feature New capabilities, enhancements, or enablement work full-test-suite needs-more-tests Requires additional L0 and L1 test coverage before merge
#6123 opened Sep 16, 2026 by casper-hansen Loading…
4 of 5 tasks
1
fix(conversion): support BF16 wire export for grouped MXFP8 experts
#6122 opened Sep 16, 2026 by seonjinn Contributor • Draft
1
feat(gemma4): allow CP > 1 on the dense model by building its layer spec with TE
#6120 opened Sep 16, 2026 by nvegesna-netizen Contributor • Draft
1
fix(peft): support mixed-dtype parallel LoRA community-request
#6119 opened Sep 16, 2026 by hershg Contributor • Draft
1
ProTip! Exclude everything labeled bug with -label:bug.