Sitelet https://github.com/NVIDIA-NeMo/Automodel/pulls
Skip to content

Pull requests: NVIDIA-NeMo/Automodel

New pull request New
207 Open 2,830 Closed
207 Open 2,830 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(nemotron-v3): align latent projection input dtype
#3657 opened Aug 25, 2026 by HuiyingLi Contributor • Draft
3 tasks done
1
1
perf(dllm): use the DeepEP dispatcher in the DiffusionGemma ep=8 recipes
#3654 opened Aug 25, 2026 by akoumpa Contributor Loading…
1
feat(dllm): add Transformer Engine context parallelism for DiffusionGemma
#3651 opened Aug 25, 2026 by akoumpa Contributor • Draft
5
test(ep): add qwen3_5_moe expert-parallel parity test [3/4]
#3649 opened Aug 24, 2026 by athitten Contributor • Draft
1
feat(registry): add public architecture registration and entry-point discovery
#3645 opened Aug 24, 2026 by pstjohn Loading…
2
fix(test): unblock Kimi-Linear HF reference and gate MiniMax routed parity
#3644 opened Aug 24, 2026 by yuhezhang-ai Contributor • 2/2 • Draft
1
fix(test): repair the MiniMax M2.7 vanilla-HF parity reference
#3643 opened Aug 24, 2026 by yuhezhang-ai Contributor • 1/2 • Draft
1
fix(moe): equalize and align per-rank token counts for HybridEP dispatch
#3641 opened Aug 24, 2026 by HuiyingLi Contributor Loading…
1
feat(laguna): support packed THD context parallelism
#3640 opened Aug 24, 2026 by akoumpa Contributor Loading…
3 tasks done
8
fix: Canonicalize LoRA compute dtype community-request
#3638 opened Aug 24, 2026 by benthecarman Loading…
3 tasks done
3
fix(glm): align and diagnose cross-framework router parity
#3635 opened Aug 22, 2026 by yuhezhang-ai Contributor • 3/3 • Draft
6
chore: bump lint tools + fix lint issues community-request
#3634 opened Aug 22, 2026 by akx Contributor Loading…
3 tasks done
2
fix(kimi_k25): honour the quantization flag community-request waiting-on-customer Waiting on the original author to respond
#3633 opened Aug 22, 2026 by akx Contributor Loading…
3 tasks done
6
fix(qwen3_omni_moe): put the thinker prefix inside the peft prefix on lora saves community-request
#3630 opened Aug 22, 2026 by stanley1208 Contributor Loading…
1
2
ci: Update transformers to latest version 5.15.1
#3628 opened Aug 22, 2026 by svcnvidia-nemo-ci Contributor Loading…
1
feat(dflash): support top-p and top-k in speculative decoding community-request waiting-on-customer Waiting on the original author to respond
#3627 opened Aug 22, 2026 by kashif Contributor Loading…
3 tasks done
11
fix(moe): export fused expert lora in the corrected peft >= 0.19.1 layout community-request
#3626 opened Aug 22, 2026 by stanley1208 Contributor Loading…
1
2
fix(utils): accumulate param L2 norm in float32 for bf16/fp16 models community-request
#3625 opened Aug 22, 2026 by ralovets Contributor Loading…
2 of 3 tasks
2
perf(checkpoint): bound GPT-OSS MXFP4 loading
#3623 opened Aug 22, 2026 by yuhezhang-ai Contributor • 5/5 • Draft
3
fix(checkpoint): avoid DCP collectives during initialization
#3622 opened Aug 22, 2026 by HuiyingLi Contributor • Draft
3 tasks done
1
perf(checkpoint): avoid full GC scans during MoE export
#3621 opened Aug 22, 2026 by yuhezhang-ai Contributor Loading…
1
test(ci): report Jensen-Shannon checkpoint parity metrics
#3620 opened Aug 22, 2026 by yuhezhang-ai Contributor • 2/3 • Draft
3 tasks done
1
perf(checkpoint): bound quantized DCP load memory
#3619 opened Aug 21, 2026 by yuhezhang-ai Contributor • 4/5 • Draft
1
fix(moe): forward permute fusion to HybridEP
#3618 opened Aug 21, 2026 by HuiyingLi Contributor • Draft
4 tasks done
1
perf(checkpoint): load standard HF safetensors with DCP
#3616 opened Aug 21, 2026 by yuhezhang-ai Contributor • 3/5 • Draft
1
ProTip! Mix and match filters to narrow down what you’re looking for.