Skip to content

Pull requests: NVIDIA/TransformerEngine

New pull request New
188 Open 2,424 Closed
188 Open 2,424 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: guard backward_dw() bias-grad backfill against a legitimately-empty bgrad community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3400 opened Aug 18, 2026 by nvegesna-netizen Contributor Loading…
1
Publishing updates community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3399 opened Aug 18, 2026 by zredeaux07 • Draft
5
[PyTorch] Type the grouped wgrad from main_grad for distributed weights org-contribution
#3397 opened Aug 18, 2026 by fanshiqing Member Loading…
8 of 13 tasks
1
2
[CI] Add workflow verifying build against minimum supported PyTorch and JAX
#3396 opened Aug 18, 2026 by pggPL Collaborator • Draft
8 of 13 tasks
2
[Common][PyTorch] Add QB router histogram paths org-contribution
#3395 opened Aug 18, 2026 by harryzhou2000 Member Loading…
8 of 13 tasks
5
[PyTorch] Pair delayed-scaling FP8 recompute metadata per module community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3394 opened Aug 18, 2026 by nvegesna-netizen Contributor Loading…
7 of 13 tasks
@pggPL
3
Support quantized tensor subclasses across Python and C++ community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3393 opened Aug 18, 2026 by dingqingy-nv Loading…
8 of 13 tasks
@ptrendx
12
[Common] Move arch specific helpers in ptx.cuh to a new header
#3392 opened Aug 17, 2026 by kainzhong Collaborator Loading…
8 of 13 tasks
@ptrendx
5
Thread tanh logit softcapping through FlashAttention (FA2, opt-in FA3) community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3391 opened Aug 17, 2026 by nvegesna-netizen Contributor Loading…
9
fix: duplicate pad_between_seqs skips and wrong architecture direction community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3390 opened Aug 17, 2026 by andrewwhitecdw Contributor Loading…
6
fix: guard softmax_offset.grad.zero_() by is_training community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3389 opened Aug 17, 2026 by andrewwhitecdw Contributor Loading…
1
fix: scan-ring env var may leak across tests on pytest.skip community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3388 opened Aug 17, 2026 by andrewwhitecdw Contributor Loading…
4
fix: reorder test uses original shape after swapping SBHD axes community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3387 opened Aug 17, 2026 by andrewwhitecdw Contributor Loading…
1
[PyTorch] Share MXFP8 data tensor between rowwise and columnwise for 2D quantization community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3385 opened Aug 17, 2026 by TangChangcheng Loading…
4
surface comm overlap rank errors
#3384 opened Aug 14, 2026 by sudhakarsingh27 Member Loading…
2
Produce BOLT-compatible libraries
#3383 opened Aug 14, 2026 by fheinecke Collaborator Loading…
2 of 13 tasks
5
[PyTorch] Fuse eager EP prepare and dispatch into a single op
#3380 opened Aug 13, 2026 by phu0ngng Collaborator • Draft
13 tasks
2
fix: duplicate JUnit XML filename overwrites encoder test results community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3376 opened Aug 13, 2026 by andrewwhitecdw Contributor Loading…
2
[JAX]: SBHD reorder skip uses original shape instead of swapped tensor community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3373 opened Aug 13, 2026 by andrewwhitecdw Contributor Loading…
11
fix: comment typo 'mas' -> 'mask' in two TODOs community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3372 opened Aug 13, 2026 by andrewwhitecdw Contributor Loading…
3
fix: make_mask uses deprecated jnp.bool instead of jnp.bool_ community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3371 opened Aug 13, 2026 by andrewwhitecdw Contributor Loading…
6
fix: Optional segment position annotations in mask helpers community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3369 opened Aug 13, 2026 by andrewwhitecdw Contributor Loading…
2
fix: Unreachable backend check after earlier backend skip community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3368 opened Aug 13, 2026 by andrewwhitecdw Contributor Loading…
4
fix: remove redundant self-assignment out_ = out_ community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3367 opened Aug 13, 2026 by andrewwhitecdw Contributor Loading…
10
fix: simplify qkv_format checks using a membership tuple community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#3365 opened Aug 13, 2026 by andrewwhitecdw Contributor Loading…
1
ProTip! Updated in the last three days: updated:>2026-08-15.