Skip to content

Pull requests: NVIDIA-NeMo/Automodel

New pull request New
98 Open 1,420 Closed
98 Open 1,420 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: fallback to safetensors if using peft
#1924 opened Apr 20, 2026 by akoumpa Contributor • Draft
3 tasks
2
fix: make _get_logits pp aware in ckpt robustness r0.4.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge.
#1923 opened Apr 20, 2026 by akoumpa Contributor • Draft
3 tasks
2
fix: llava onevision recipes
#1922 opened Apr 20, 2026 by akoumpa Contributor • Draft
3 tasks
2
fix: chat dataset r0.4.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge.
#1921 opened Apr 20, 2026 by akoumpa Contributor Loading…
3 tasks
3
cp: fix: qlora ckpt loading (1549) into r0.4.0 cherry-pick Run CICD Trigger Testing CICD
#1920 opened Apr 20, 2026 by svcnvidia-nemo-ci Contributor Loading…
2
fix: update defer_fsdp_grad_sync in recipes r0.4.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge.
#1919 opened Apr 20, 2026 by akoumpa Contributor • Draft
3 tasks
5
fix: Step-3.5-Flash layer_types mismatch and related recipe fixes r0.4.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge.
#1916 opened Apr 20, 2026 by hemildesai Contributor Loading…
2 of 3 tasks
9
ci: Add test_recipes for custom test scope docs-only With great power comes great responsibility. r0.4.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge.
#1915 opened Apr 20, 2026 by thomasdhc Contributor • Draft
3 tasks
3
feat: add Context Parallelism support for Gemma4 dense and MoE VLM community-request
#1914 opened Apr 20, 2026 by khazic Contributor Loading…
3 tasks done
9
fix(moe): align EP expert weight dtype with activation dtype
#1913 opened Apr 20, 2026 by jQizhang Contributor Loading…
1 of 3 tasks
@hemildesai
4
fix: fp32 master weights for custom MoE models under FSDP2
#1896 opened Apr 17, 2026 by zpqiu Contributor Loading…
1 of 3 tasks
6
ci: onboard GB200 testing
#1893 opened Apr 17, 2026 by ko3n1g Contributor Loading…
3 tasks
ci(feat): use AWS ephemeral runners for external contributors
#1892 opened Apr 17, 2026 by ko3n1g Contributor Loading…
2 tasks
1
fix: lora with gemma4 large models on Spark single GPU
#1866 opened Apr 15, 2026 by athitten Contributor • Draft
3 tasks
1
5
feat: Qwen3.5 VLM TP+PP support with per-microbatch grad reduce-scatter knob r0.4.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge.
#1859 opened Apr 15, 2026 by akoumpa Contributor Loading…
6 tasks done
5
docs: add embedding + reranker model coverage docs-only With great power comes great responsibility.
#1843 opened Apr 14, 2026 by akoumpa Contributor Loading…
3 tasks
36
ci: add sync-skills workflow
#1841 opened Apr 14, 2026 by ko3n1g Contributor Loading…
2 tasks
3
feat: add extract_submodel parameter to build_encoder_backbone
#1838 opened Apr 14, 2026 by oliverholworthy Contributor • Draft
2 of 3 tasks
3
refactor: use config.is_causal=False for bidirectional attention
#1837 opened Apr 14, 2026 by oliverholworthy Contributor • Draft
2 of 3 tasks
2
refactor: Remove separate moe_mesh references community-request waiting-on-customer Waiting on the original author to respond
#1824 opened Apr 14, 2026 by edjson Loading…
3 tasks
1
3
ci: Update transformers to latest version 5.5.4
#1823 opened Apr 14, 2026 by svcnvidia-nemo-ci Contributor Loading…
1
feat: Add Ministral3-3B bidirectional encoder training scripts
#1809 opened Apr 13, 2026 by rnyak Collaborator • Draft
3 tasks
2
fix: Set CUDA arch list for UCCL EP build to SM90+ r0.4.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge.
#1808 opened Apr 13, 2026 by thomasdhc Contributor Loading…
3 tasks
ci: add HF doc contract test for pretraining instantiation r0.4.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge.
#1802 opened Apr 13, 2026 by adil-a Collaborator • Draft
3 tasks done
1
test: add dataloader checkpoint integration test for retrieval recipes
#1800 opened Apr 13, 2026 by oliverholworthy Contributor Loading…
2 of 3 tasks
@oliverholworthy
ProTip! no:milestone will show everything without a milestone.