Skip to content

Pull requests: NVIDIA/TransformerEngine

New pull request New
80 Open 1,523 Closed
80 Open 1,523 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] fix cross entropy vanishing gradients
#2139 opened Aug 29, 2025 by casper-hansen Loading…
1 of 13 tasks
1
ci: Build and attach bdist wheels to release page
#2138 opened Aug 29, 2025 by ko3n1g Loading…
13 tasks
[PyTorch Debug] Add max_blockwise_X_dynamic_range stats
#2137 opened Aug 29, 2025 by pggPL Loading…
8 of 13 tasks
[JAX] HighPrecisionTensor wrapper for non-quantized data
#2136 opened Aug 28, 2025 by jberchtold-nvidia Loading…
8 of 13 tasks
2
[JAX] Fix failing fused attn tests for dropout=0.1 and bias for sm100
#2135 opened Aug 28, 2025 by KshitijLakhani Loading…
5 of 13 tasks
@KshitijLakhani
2
Fix CUDA version in setup.py
#2132 opened Aug 27, 2025 by vcherepanov-nv • Draft
1 of 13 tasks
1
Create GPU reload buffers on main stream
#2131 opened Aug 27, 2025 by sanandaraj5597 Loading…
Adds context parallelism utilities: moving cp shards to diff ranks and pad sequence to divisibility factory 2.8.0
#2129 opened Aug 27, 2025 by jomitchellnv Loading…
5 of 12 tasks
Fix memory overhead of linear layer when all gather from sequence parallel
#2125 opened Aug 27, 2025 by yuzhongw-nvidia Loading…
13 tasks
13
FP8 Output Quantization for GEMM
#2123 opened Aug 26, 2025 by vthumbe1503 • Draft
6 of 13 tasks
Fused RoPE with combined QKV input.
#2122 opened Aug 26, 2025 by vasunvidia Loading…
1 of 13 tasks
5
Adds dst.dtype information in copy_ method of quantized tensors.
#2120 opened Aug 26, 2025 by zobeideThePlayer Loading…
3 of 13 tasks
4
[PyTorch][CUDA Graph] Fix FP8 Weight Quantization Cache under CUDA Graph performance Performance issues
#2119 opened Aug 26, 2025 by zhongbozhu Loading…
13 tasks
@zhongbozhu
5
[PyTorch Debug] Draft of tutorial for custom precision testing.
#2112 opened Aug 25, 2025 by pggPL • Draft
1 of 13 tasks
[PyTorch Debug] Fix issue with microbatching + debug value caching
#2108 opened Aug 25, 2025 by pggPL Loading…
8 of 13 tasks
2
[PyTorch Debug] Fix issue with negative underflow.
#2107 opened Aug 25, 2025 by pggPL Loading…
8 of 13 tasks
11
Fix test of FSDP2 by correcting init logic and applying autocast
#2105 opened Aug 24, 2025 by ntenenz Loading…
4 of 13 tasks
1
4
[PyTorch] Fix assertion error message formatting in DotProductAttention
#2103 opened Aug 22, 2025 by janbernloehr Loading…
6 of 13 tasks
2
[PyTorch] Let GroupedLinear accept MXFP8 input and gradient
#2099 opened Aug 21, 2025 by yaox12 Loading…
1 of 13 tasks
8
Support communication/gemm overlap for [Wgrad->Dgrad] execution order in the bwd pass.
#2065 opened Aug 12, 2025 by fanshiqing Loading…
3 of 13 tasks
1
Support MLA Context Parallel (CP) exchanging latent KV
#2064 opened Aug 12, 2025 by yuzhongw-nvidia Loading…
13 tasks
Feature fast cast-only mxfp8
#2062 opened Aug 12, 2025 by Jianbing-D Loading…
6 of 13 tasks
16
Support PyPI wheel for cuda13 build Build system
#2057 opened Aug 11, 2025 by ksivaman Loading…
2 of 13 tasks
@timmoon10
5
Add better ordering enforcment to split_overlap_rs gemms.
#2056 opened Aug 11, 2025 by chaseblock Loading…
6 of 13 tasks
2
[Draft] Add primary weighs fp8 support for mxfp8
#2055 opened Aug 11, 2025 by kunlunl Loading…
13 tasks
ProTip! Exclude everything labeled bug with -label:bug.