Skip to content

Pull requests: NVIDIA/TransformerEngine

New pull request New
123 Open 2,020 Closed
123 Open 2,020 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

Bump transformers from 4.57.0 to 5.0.0rc3 in /docs/examples/te_llama dependencies Pull requests that update a dependency file python Pull requests that update python code
#2850 opened Apr 8, 2026 by dependabot bot Loading…
Simplify FA3 discovery
#2849 opened Apr 8, 2026 by vcherepanov-nv Loading…
4 of 13 tasks
2
Skip activation kernels when tensor size is zero bug Something isn't working
#2848 opened Apr 8, 2026 by timmoon10 Loading…
8 of 13 tasks
2
[Common] Multicast Fixes
#2847 opened Apr 8, 2026 by phu0ngng • Draft
13 tasks
1
Add Megatron-FSDP E2E integration test to TE CI/CD (L1).
#2845 opened Apr 7, 2026 by cspades Loading…
3 of 13 tasks
8
[Core] Report CUDA versions when NVRTC compilation fails enhancement New feature or request
#2842 opened Apr 7, 2026 by timmoon10 Loading…
8 of 13 tasks
1
13
comm_gemm_test fixes
#2839 opened Apr 6, 2026 by almogsegal Loading…
13 tasks
@denera
6
[CUDNN] Update frontend to version 1.22 and add cuDNN 9.20 path for SM arch >100
#2838 opened Apr 5, 2026 by zmelumian972 Loading…
2 of 3 tasks
3
Add grouped unswizzle functionality for MXFP8 scaling factors
#2837 opened Apr 5, 2026 by int-smart Loading…
8 of 13 tasks
@ptrendx
6
[PyTorch] Fix FlashAttention 2 head_dim > 192 on sm103 and other architectures
#2836 opened Apr 4, 2026 by pedramr Loading…
1 task done
3
Fix JAX extension build with NVTE_UB_WITH_MPI=1
#2835 opened Apr 4, 2026 by GaetanLepage Loading…
2 of 13 tasks
2
[Pyt][Common] Enabling/Guarding sm120 support (non - attention)
#2833 opened Apr 3, 2026 by KshitijLakhani • Draft
13 tasks
@KshitijLakhani
fix CUDA architectures cmake logic community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#2832 opened Apr 3, 2026 by GaetanLepage Loading…
2 of 13 tasks
4
Add capture_time_hooks to make_graphed_callables for non-capturable per-callable hooks
#2831 opened Apr 3, 2026 by buptzyb Loading…
1 of 13 tasks
17
Port softmax ops to libtorch stable ABI
#2830 opened Apr 3, 2026 by pstjohn Loading…
22
Cp thd swa with ag
#2829 opened Apr 3, 2026 by sudhakarsingh27 • Draft
13 tasks
3
[Common] Reduced padding kernel compilation time
#2827 opened Apr 2, 2026 by Oleg-Goncharov Loading…
5 of 13 tasks
1
fix(CP, MLA): CP works fine with MLA in a2a cp_comm_type community-contribution PRs from external contributor outside the core maintainers, representing community-driven work.
#2826 opened Apr 2, 2026 by zhujian19891203 Loading…
5 of 13 tasks
2
fix(CP, FA): the conditional logic in the FA version contains a vulnerability when processing the output of Flash Attn forward pass
#2825 opened Apr 2, 2026 by zhujian19891203 Loading…
5 of 13 tasks
1
Parallel Test Execution to decrease CI run times
#2824 opened Apr 2, 2026 by sudhakarsingh27 • Draft
[Common] Fix fused router for large top-K and expert counts
#2821 opened Apr 1, 2026 by harryzhou2000 Loading…
7 of 13 tasks
@denera
15
[Pytorch][Common] Hybrid quantization
#2817 opened Mar 31, 2026 by negvet Loading…
1 of 13 tasks
9
[PyTorch] [torch.compile] Remove internal tensor state from Float8CurrentScalingQuantizer
#2816 opened Mar 31, 2026 by pggPL • Draft
13 tasks
Add PyTorch stable ABI extension for libtorch compatibility
#2813 opened Mar 30, 2026 by pstjohn • Draft
6
[PyTorch] [torch.compile] Split linear forward into forward and setup context.
#2811 opened Mar 30, 2026 by pggPL Loading…
8 of 13 tasks
5
ProTip! Add no:assignee to see everything that’s not assigned.