Skip to content

Pull requests: NVIDIA-NeMo/Automodel

New pull request New
172 Open 2,306 Closed
172 Open 2,306 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(distributed): support packed CP for Llama, Qwen2, and Qwen3
#2999 opened Jul 9, 2026 by akoumpa Contributor • Draft
3 tasks done
1
refactor(models): single tie_word_embeddings guard via TieSupport community-request
#2998 opened Jul 9, 2026 by Achyuthan-S Contributor Loading…
1 of 11 tasks
1
6
fix(optim): drop zero-numel local DTensor shards before TE FusedAdam
#2997 opened Jul 9, 2026 by yuhezhang-ai Contributor • Draft
2
feat(loss): chunked masked CE, packing-mask preservation, scoped MoE AC
#2996 opened Jul 9, 2026 by yuhezhang-ai Contributor • Draft
3
fix(moe): safe TP and EP/TP gradient correctness for custom MoE models
#2995 opened Jul 9, 2026 by yuhezhang-ai Contributor • Draft
3
ci(review): require random-init invariant checks docs-only With great power comes great responsibility.
#2994 opened Jul 9, 2026 by akoumpa Contributor • Draft
3
fix(moe): checkpoint trainable vision towers on the expert-parallel path
#2993 opened Jul 9, 2026 by yuhezhang-ai Contributor • Draft
3
feat(training): opt-in setup-time prewarms + dry-run warmup RFC
#2992 opened Jul 9, 2026 by yuhezhang-ai Contributor • Draft
2
fix(deepseek-v4): initialize random training state
#2991 opened Jul 9, 2026 by akoumpa Contributor Loading…
4
feat(distributed): frame-level context-parallel vision-tower sharding
#2990 opened Jul 9, 2026 by yuhezhang-ai Contributor • Draft
3
feat(distributed): block-diagonal varlen CP for packed sequences
#2989 opened Jul 9, 2026 by yuhezhang-ai Contributor • Draft
2
fix: raise CI wall-clock for 4 release recipes timing out (AM-629) ci docs-only With great power comes great responsibility. Run CICD Trigger Testing CICD
#2988 opened Jul 9, 2026 by svcnemo-autobot Loading…
2
feat(distributed): shape TPLinear/LinearLoRA graphs for async-TP fusion
#2987 opened Jul 9, 2026 by yuhezhang-ai Contributor • Draft
2
fix(distributed): Megatron-FSDP 0.5.0 compatibility
#2986 opened Jul 9, 2026 by yuhezhang-ai Contributor • Draft
3
perf(Gemma4): reduce comms in sliding window, FA implementation, community-request
#2984 opened Jul 9, 2026 by nathan-az • Draft
3 tasks
3
build(deps): bump base container to 26.06-cuda13.3
#2983 opened Jul 9, 2026 by thomasdhc Contributor Loading…
3 tasks
1
fix(attention): initialize FlexAttention masks
#2982 opened Jul 9, 2026 by akoumpa Contributor Loading…
3 tasks done
4
feat: Support for embedding/retriever model distillation
#2981 opened Jul 9, 2026 by vinay-raman Loading…
3 tasks done
1
ci(review): adopt hardened reusable workflow
#2980 opened Jul 9, 2026 by akoumpa Contributor • Draft
1
perf: chunk MoE per-expert loop by routed tokens community-request
#2979 opened Jul 9, 2026 by mohamed-alansary Loading…
2
chore: CI proxy for #2966
#2978 opened Jul 9, 2026 by HuiyingLi Contributor Loading…
2
perf: use input dtype for MoE scatter accumulator under no_grad community-request
#2977 opened Jul 9, 2026 by mohamed-alansary Loading…
1
feat(speculative): FSDP2-shard dense DSpark target + add Gemma4-31B config community-request
#2976 opened Jul 9, 2026 by khazic Contributor Loading…
11
perf: chunk grouped-mm scatter source conversion community-request
#2975 opened Jul 9, 2026 by mohamed-alansary Loading…
1
fix: chunk variable-length all-gather backward community-request
#2974 opened Jul 9, 2026 by mohamed-alansary Loading…
4
ProTip! What’s not been updated in a month: updated:<2026-06-09.