Skip to content

Pull requests: NVIDIA/TransformerEngine

New pull request New
70 Open 1,412 Closed
70 Open 1,412 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

[C][PyTorch] Remove deprecated device_id arg for multi tensor API
#1994 opened Jul 24, 2025 by ksivaman Loading…
6 of 13 tasks
2
[JAX] Temporary disable Norm custom calls
#1993 opened Jul 24, 2025 by phu0ngng • Draft
13 tasks
[PyTorch] Optimize cudagraph static_grad_outputs reuse
#1992 opened Jul 24, 2025 by buptzyb Loading…
5 of 13 tasks
Fixed double buffering issue for assymetric layers
#1984 opened Jul 23, 2025 by sanandaraj5597 Loading…
Use userbuffers for MXFP8 wgrad all-gather overlap
#1982 opened Jul 22, 2025 by djns99 • Draft
1 of 13 tasks
[JAX] Fixing GemmPrimitive partitioning rules to handle tensor-parallelism correctly for sequence-parallel inputs 2.6.0
#1980 opened Jul 22, 2025 by denera Loading…
8 of 13 tasks
@denera
7
[PyTorch] fix input_quantizer usage in Linear backward for save_original_input
#1978 opened Jul 22, 2025 by hxbai Loading…
8 of 13 tasks
1
Rename do_not_clear to _do_not_clear
#1977 opened Jul 21, 2025 by janekb04 Loading…
3 of 13 tasks
1
Manually launch wgrad accumulation and reduce in backward_dw() instead of backward()
#1976 opened Jul 21, 2025 by lhb8125 Loading…
2 of 13 tasks
6
[C][PyTorch] NVFP4 forward MXFP8 backward recipe
#1970 opened Jul 18, 2025 by ksivaman • Draft
2 of 13 tasks
10
[PyTorch] Enable generic QK norm support (+ RMSNorm/LayerNorm)
#1966 opened Jul 18, 2025 by negvet Loading…
7 of 13 tasks
1
7
[PyTorch][Mcore] Fix illegal memory access issue while using Mcore async checkpoint with fp8 tensorwise recipe bug Something isn't working
#1956 opened Jul 16, 2025 by zhongbozhu Loading…
13 tasks
2
[PyTorch Debug] Debug support for GroupedLinear
#1953 opened Jul 15, 2025 by pggPL • Draft
13 tasks
[PyTorch] Refactor C++ quantizer infrastructure
#1952 opened Jul 15, 2025 by timmoon10 Loading…
8 of 13 tasks
11
[PyTorch][FP8 CS] Remove the unnecessary torch reciprocal op in fp8 current scaling code path performance Performance issues
#1950 opened Jul 14, 2025 by zhongbozhu Loading…
13 tasks
@zhongbozhu
10
[PyTorch] Support delay_wgrad_compute cudagraph
#1948 opened Jul 14, 2025 by buptzyb Loading…
2 of 13 tasks
[Minor] Update 1_getting_started.rst documentation Improvements or additions to documentation
#1947 opened Jul 13, 2025 by dupeljan Loading…
3 of 8 tasks
2
[JAX] Select cuDNN backend for normalization
#1946 opened Jul 11, 2025 by phu0ngng Loading…
13 tasks
7
[BUILD] Exclude ninja from required packages
#1932 opened Jul 7, 2025 by phu0ngng Loading…
5 of 13 tasks
4
[PyTorch] Fuse permute+pad and unpermute+unpad ops for FP8 optimization
#1921 opened Jul 3, 2025 by xiaoxi-wangfj Loading…
3 of 12 tasks
3
Fix import error when flash attention 3 is installed
#1913 opened Jun 30, 2025 by HollowMan6 Loading…
7 of 13 tasks
[PyTorch debug] Improve precision debug tools performance
#1909 opened Jun 30, 2025 by pggPL Loading…
9 of 13 tasks
5
[PyTorch] Support FA3 MLA CP feature
#1907 opened Jun 28, 2025 by zhujian19891203 Loading…
7 of 13 tasks
@cyanguwa
[PyTorch Debug] Support log fp8 tensor stats for blockwise recipe
#1905 opened Jun 27, 2025 by lengerfulluse Loading…
12 tasks
[Common] NVFP4 kernels enhancement New feature or request
#1904 opened Jun 27, 2025 by Oleg-Goncharov • Draft
5 of 13 tasks
ProTip! Follow long discussions with comments:>50.