Skip to content

Pull requests: NVIDIA/TransformerEngine

New pull request New
116 Open 1,938 Closed
116 Open 1,938 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] Add QuantizedTensor support in FusedAdam.step for MXFP8BlockScaling and Float8BlockScaling quantized model init.
#2753 opened Mar 11, 2026 by jomitchellnv Loading…
13 tasks
15
[JAX] Change dtype of intermediate result aval of fused_topk_and_score_function_fwd to fp32
#2752 opened Mar 10, 2026 by tdophung Loading…
1 of 13 tasks
3
Support configurable number of philox rounds for stochastic rounding 2.14.0
#2751 opened Mar 10, 2026 by ksivaman Loading…
7 of 13 tasks
3
[PyTorch] Fix fuser so it releases tensors properly
#2750 opened Mar 10, 2026 by kainzhong Loading…
5 of 13 tasks
6
[JAX] Grouped GEMM Refactor to use first_dims and last_dims
#2749 opened Mar 10, 2026 by jberchtold-nvidia • Draft
1 of 13 tasks
7
[Core] MXFP8 grouped GEMM + tensor-scaled FP8 fixes
#2748 opened Mar 9, 2026 by jberchtold-nvidia Loading…
13 tasks
18
[JAX] Add bias support for v2 grouped GEMM path
#2744 opened Mar 6, 2026 by jberchtold-nvidia Loading…
8 of 13 tasks
4
[Common] Persistent Grouped NVFP4 quantization kernel
#2743 opened Mar 6, 2026 by Oleg-Goncharov Loading…
8 of 13 tasks
30
Add guard at lowest JAX version that still supports triton kernel calling
#2741 opened Mar 6, 2026 by tdophung Loading…
6 of 13 tasks
33
[JAX] Collective GEMM with FP8 and MXFP8 support
#2740 opened Mar 5, 2026 by phu0ngng Loading…
7 of 13 tasks
5
[Common] Persistent Grouped MXFP8 quantization kernel enhancement New feature or request MoE
#2738 opened Mar 5, 2026 by Oleg-Goncharov Loading…
9 of 13 tasks
15
Feat/cp nvshmem enhanced community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#2737 opened Mar 5, 2026 by Knight-of-Thunder Loading…
1 of 13 tasks
12
Feature/unswizzle community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#2732 opened Mar 4, 2026 by int-smart Loading…
9 of 13 tasks
16
fix: scope get_full_cu_seqlens cache key by device and inference mode
#2728 opened Mar 3, 2026 by DmCarpe93 Loading…
8 of 13 tasks
3
[CI] Refactor CI build on GitHub
#2723 opened Mar 2, 2026 by ptrendx Loading…
1 of 13 tasks
4
[Common, pyTorch] Grouped MXFP8 dequantize support
#2722 opened Mar 2, 2026 by ptrendx Loading…
1 of 13 tasks
1
2
Fix for async dcp checkpointing with Float8Tensors
#2721 opened Mar 2, 2026 by pstjohn • Draft
Add MXFP8 attention
#2719 opened Mar 1, 2026 by cyanguwa • Draft
13 tasks
10
[PyTorch] Fix CPU offloading for bulk-allocated quantized tensors
#2716 opened Feb 27, 2026 by lhb8125 • Draft
1 of 13 tasks
6
Add DCP compatibility for FSDP2-TP sharding in TransformerEngine.
#2713 opened Feb 26, 2026 by cspades Loading…
3 of 13 tasks
14
[Common][PyTorch] Add z_loss_weight and log_sum_exp output to parallel_cross_entropy
#2707 opened Feb 26, 2026 by bassoy • Draft
8 tasks done
Newton-Schulz via cuSOLVERMp
#2706 opened Feb 25, 2026 by vcherepanov-nv Loading…
6 of 13 tasks
28
[All] Added better error messages
#2705 opened Feb 25, 2026 by ptrendx Loading…
15
[Draft][PyTorch] torch.compile support for TE Linear
#2701 opened Feb 24, 2026 by pggPL • Draft
13 tasks
Enable sm120 support for fused attn if cuDNN is 9.18.1+
#2693 opened Feb 20, 2026 by KshitijLakhani • Draft
13 tasks
ProTip! Add no:assignee to see everything that’s not assigned.