COMMITS
August 26, 2026
M
[fx] Clear fake_mode on a view's base in GraphPickler (#194771) (#194771)
Michael Roberts committed
W
[CI] Install torchvision for the torchtitan integration tests (#194702)
Wei Feng committed
P
Revert "Move c10/util math helper headers into torch/headeronly (#194673)"
PyTorch MergeBot committed
C
Move c10/util math helper headers into torch/headeronly (#194673)
Chris Leonard committed
J
E
Remove dead mixed_mm remnants from inductor and torchgen (#193721)
Egemen Tuncarslan committed
August 25, 2026
J
[dynamo] clarify verbose hint for backend failures (#185419)
Jason Ansel committed
E
[inductor] Fix loop-local load CSE lifetime (#194786)
eellison committed
S
A
B
H
Remove the test retry policy A/B experiment from unstable (#194779)
Huy Do committed
J
Fix Inductor bool vector mask casts (#184155)
Jason Ansel committed
Y
Add per-operator headers to XPU sources (#194688)
Yuanyuan Chen committed
J
Fix AOT export GraphModule serialization (#186418)
Jason Ansel committed
A
Bump the Python 3.15 numpy pin to 2.5.2 (#194741)
Andrey Talman committed
N
[BE] MacOS CD: Update `actions/` to latest majors (#194748)
Nikita Shulga committed
A
Fix docker-release validate job to use the channel matching the pushed image (#194788)
Andrey Talman committed
S
[inductor] Honor assume_32bit_indexing in _decide_tl_dtype (#194127) (#194127)
Simran Barnwal committed
W
[optim][muon] fix momentum buffer aliasing in Newton-Schulz (#194664)
Wei Feng committed
Z
A
June 17, 2026
J
Handle invalid CUDA JITerator cache entries (#186346)
Jason Ansel committed
August 24, 2026
F
Fix typos and grammar in comments and docstrings (#194567)
Frederik Gossen committed
F
Fix grammar in comments and docstrings (#194624)
Frederik Gossen committed
F
Fix typos in comments and docstrings (#194627)
Frederik Gossen committed
F
Fix grammatical errors in comments and docstrings (#194564)
Frederik Gossen committed
August 23, 2026
A
Strict pyrefly checking: type torch/_numpy/_funcs_impl.py (#194529)
Aaron Orenstein committed
August 21, 2026
J
Simplify stateless RNG split() / fold_in() kernels with no grid-stride loop (#190036)
Joel Schlosser committed