Skip to content

Pull requests: NVIDIA/TransformerEngine

New pull request New
109 Open 1,814 Closed
109 Open 1,814 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

[PyTorch] ONNX test fix + export for FP8 attention
#2598 opened Jan 14, 2026 by pggPL • Draft
2 of 13 tasks
1
[PyTorch] Support user-defined op fusions enhancement New feature or request
#2597 opened Jan 14, 2026 by timmoon10 Loading…
7 of 13 tasks
5
Flash attn pad bw seqs
#2596 opened Jan 14, 2026 by sudhakarsingh27 Loading…
2 of 7 tasks
1
@sudhakarsingh27
3
fix: enable opt for cutlass sources to avoid infinite compile time build Build system
#2595 opened Jan 14, 2026 by kainzhong Loading…
5 of 13 tasks
2
Initial commit to pass scale as Tensor for multi_tensor_scale op
#2594 opened Jan 13, 2026 by vasunvidia • Draft
13 tasks
[JAX] HLO FFI tests
#2593 opened Jan 13, 2026 by jberchtold-nvidia • Draft
7 of 13 tasks
7
Revert adding pytorch-triton as a build requirement 2.12.0
#2592 opened Jan 13, 2026 by tdophung Loading…
5 of 13 tasks
3
[Draft] [JAX] Custom partitioning for Permutation primitives MoE
#2591 opened Jan 13, 2026 by tdophung Loading…
13 tasks
4
(Bug fix) Fix accuracy issue for blockwise scaling+E8 scale on Blackwell
#2589 opened Jan 13, 2026 by lhb8125 Loading…
13 tasks
23
[Common] MXFP8 kernel for grouped tensors
#2586 opened Jan 12, 2026 by Oleg-Goncharov • Draft
13 tasks
[Common] Enable determinism for cuDNN >= 9.18 on Blackwell 2.12.0
#2584 opened Jan 12, 2026 by cyanguwa Loading…
8 of 13 tasks
52
docs: Update README Latest News section
#2583 opened Jan 9, 2026 by sbhavani Loading…
3 of 13 tasks
1
Make router_fusion to adapt for the large num_of_expert(>2048)
#2582 opened Jan 9, 2026 by Autumn1998 Loading…
13 tasks
11
fix(build): Handle namespace packages for PyPI CUDA detection
#2580 opened Jan 9, 2026 by sbhavani Loading…
6 of 13 tasks
1
4
fix(examples): te_llama compatibility with transformers >= 4.57
#2572 opened Jan 7, 2026 by sbhavani Loading…
6 of 13 tasks
1
8
[PyT] Update THD sink attention logic for cudnn >=9.18.0 2.12.0
#2568 opened Jan 6, 2026 by cuichenx Loading…
13 tasks
11
CPU Optimizations for FP8 cpu_overhead
#2559 opened Jan 5, 2026 by vthumbe1503 Loading…
13 tasks
107
[NVFP4][Dense/MoE] Integrate Cutlass NVFP4 Row-Cast-Col-RHT-Transpose-Cast Fusion Kernel fp4 MoE
#2555 opened Jan 3, 2026 by zhongbozhu Loading…
3 of 16 tasks
@zhongbozhu
14
[Pytorch] Enhance bf16 precision optimizer performance with memory buffer
#2551 opened Dec 31, 2025 by Baidu-AIAK Loading…
4
[PyTorch] Remove unnecessary save of weights
#2549 opened Dec 30, 2025 by pggPL Loading…
8 of 13 tasks
22
[PyTorch]Add Casting-Free FP8-Flow-MoE Blockwise Optimizations community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#2544 opened Dec 26, 2025 by xiaoxi-wangfj Loading…
4 of 13 tasks
9
[PyT] Plumbing correct bias dims from TE to cudnn attention bug Something isn't working pytorch
#2537 opened Dec 20, 2025 by KshitijLakhani Loading…
5 of 11 tasks
@KshitijLakhani
13
[DO NOT MERGE] Get seqlens and offsets in O(N) space instead of O(N*N) space do not merge
#2530 opened Dec 17, 2025 by KshitijLakhani • Draft
13 tasks
@KshitijLakhani
[JAX] Calculate seqlens and offsets in O(N) space instead of O(N*N) space for THD sequences attention
#2522 opened Dec 16, 2025 by KshitijLakhani • Draft
13 tasks
@KshitijLakhani
Enable post-RHT amax estimation with separate amax scale kernel fp4
#2521 opened Dec 16, 2025 by negvet • Draft
13 tasks
1
3
ProTip! Filter pull requests by the default branch with base:main.