-
Notifications
You must be signed in to change notification settings - Fork 100
Pull requests: NVIDIA/cosmos-framework
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Upgrade pinned hf CLI to 1.26.0 for Xet-backed repos
#163
opened Aug 5, 2026 by
Sullivan07043
Loading…
Fix gradient accumulation under FSDP2: defer cross-replica all-reduce
#162
opened Aug 5, 2026 by
Sullivan07043
Loading…
feat: load ModelOpt static-FP8 checkpoints via TorchAO
#157
opened Aug 4, 2026 by
pengcuo
Collaborator
Loading…
feat(vlm): LoRA SFT support for VLM + VideoPhy-2 LoRA recipes
#155
opened Aug 4, 2026 by
lfengad
Collaborator
Loading…
fix(inference): materialise checkpoint tensors in host memory before H2D copy (17.7x faster load on GB300)
#150
opened Aug 1, 2026 by
rwagwani
Collaborator
Loading…
perf(inference): skip the generation tower and video guardrail for reasoner-only runs
#146
opened Jul 30, 2026 by
lfengad
Collaborator
Loading…
fix(inference): keep transfer attention dense for single samples
#133
opened Jul 24, 2026 by
trungtpham
Contributor
Loading…
1 of 4 tasks
feat: add compiled_region option to torch.compile the VAE encoder/decoder
#120
opened Jul 21, 2026 by
xbcReal
Loading…
Make SFT metadata filters configurable for short-task datasets
#115
opened Jul 16, 2026 by
hans-yang-tps
Contributor
•
Draft
docs(skills): add Cosmos transfer inference skill
#74
opened Jul 1, 2026 by
chrisvoncsefalvay
Contributor
Loading…
Single-GPU CPU offloading via reasoner/generator split
#11
opened Jun 2, 2026 by
rahul-steiger-nv
Loading…
ProTip!
Follow long discussions with comments:>50.