-
Notifications
You must be signed in to change notification settings - Fork 99
Pull requests: NVIDIA-NeMo/Automodel
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
perf: simplify Qwen3.5-MoE state_dict_adapter + DTensor passthrough
#1589
opened Mar 21, 2026 by
HuiyingLi
Loading…
3 tasks done
feat: GPT-OSS 20B and Moonlight 16B convergence results
#1577
opened Mar 19, 2026 by
hemildesai
Loading…
8 tasks done
fix: convert DTensor biases to local in MoE _forward_loop
#1565
opened Mar 17, 2026 by
hemildesai
Loading…
1 of 2 tasks
fix: NaN loss in NemotronH from-scratch pretraining with FSDP2
community-request
#1527
opened Mar 11, 2026 by
chloechiaw
Loading…
fix: skip initialize_weights for all NemotronH variants (including MoE)
#1526
opened Mar 11, 2026 by
terrykong
Loading…
3 tasks
fix: skip model.to(device) after checkpoint loading (tied params + FSDP)
#1489
opened Mar 8, 2026 by
terrykong
Loading…
1 of 2 tasks
fix: cherry-pick combined projection fixes (#1324, #1357) into r0.2.1
#1388
opened Feb 25, 2026 by
HuiyingLi
Loading…
2 tasks
feat: Retriever Model Support for Hybrid architecture Backbones
#1342
opened Feb 19, 2026 by
vinay-raman
•
Draft
1 of 3 tasks
Previous Next
ProTip!
Updated in the last three days: updated:>2026-03-18.