Morph MORPH
SIGN IN SIGN UP
dhruvhead / pytorch UNCLAIMED
0 0 396 Python

COMMITS

main
September 29, 2026
J
[Inductor] Don't flatten Blackwell WS template loops when Triton's Meta WS knob is on (#199061)
Janani Sriram committed
1d40e77
N
[BE] Allow disabling USE_MIMALLOC on Windows and AArch64 (#199064)
Nikita Shulga committed
e4112a1
J
Fix NumPy ndarray operators under default device contexts (#185462)
Jason Ansel committed
75f36ef
J
Delay RNN graph break until module call (#186614)
Jason Ansel committed
d5424d7
J
[dynamo] Show carets in graph break user stacks (#185466)
Jason Ansel committed
f91e138
J
Preserve CIA ops in AOT joint export (#184052)
Jason Ansel committed
3e6b629
October 1, 2026
X
[pytorch][bugfix] Add histogramdd Meta implementations (#198897) (#198897)
Xin Tong committed
582f09a
I
[MPS] Speed up QR with blocked householder updates (#198732)
Isalia20 committed
4d19930
P
[vision hash update] update the pinned vision hash (#198968)
PyTorch UpdateBot committed
5b87ef4
Y
[Inductor] Preserve input dtype in the CPU bmm mul+sum decomposition (#191324)
Yufeng He committed
198f176
P
Revert "Make Fsspec filesystem checkpointing public and exposing cache_staged_state_dict to enable async save (#188698)"
PyTorch MergeBot committed
67148e2
B
[inductor] Compile combo-kernel benchmark candidates in the async-compile pool (#198881) (#198881)
Bin Bao committed
01949e9
Z
[ROCm] Enable HSA_ENABLE_IPC_MODE_LEGACY to work around driver handle freeing issue (#198921)
Z Liu committed
77dccac
E
[dynamo] Support math.sumprod with non-constant arguments (#198685)
Eugenio "Jay" Zuccarelli committed
22669c3
I
[jit] Resolve module annotations in MRO order (#199074)
Itamar Oren committed
af63286
Y
[AI Codemod][PerfAICT] Reuse batch-dim views in bmm_out_or_baddbmm_ to cut select dispatch overhead (#198570)
Yedidya Feldblum committed
85d04a6
September 30, 2026
J
[Inductor] Count only compiled choices toward `max_template_fusion_benchmarked_choices` (#199113)
Janani Sriram committed
36a88a5
A
Avoid inspect.getattr_static when wrapping a module in OptimizedModule (#197191)
Aaron Orenstein committed
617a6f7
A
[aotautograd] Cache graphs containing autocast/inference_mode and torch._refs.tensor nodes (#191409)
Aaron Orenstein committed
070e6e3
J
Fix SingletonInt static guard evaluation (#184628)
Jason Ansel committed
462c7e0
C
[Pipeline] Enforce an outstanding send budget (#196727)
Chien-Chin Huang committed
38cca96
L
[FX splitter] Make fusion discovery deterministic (#198684) (#198684)
Limin Tang committed
bc3ce28
S
[c10d] Batch DDP reducer bucket-to-grad copy-out with _foreach_copy_ (#190524)
Sidharth Rajmohan committed
8fa740a
A
[Elastic][Distributed] Add configurable randomized backoff after failed rendezvous state writes (#185930)
alexqdh committed
bb63328
J
Prevent saving mutated grad inputs in AOT min-cut (#185876)
Jason Ansel committed
0242c7b
J
Fix symbolic stride-order sorting without explicit ShapeEnv (#185871)
Jason Ansel committed
31f0501
B
[fbgemm] Add embedding-bag byte cache re-key regression test (#199060)
Benson Ma committed
51b4a5c
S
[docs] Document undefined behavior of non-0/1 bytes in bool tensors (#199012)
Saqib Khan committed
2c857e9
H
Move the MPS M1 test shard off macOS 14 (#199118)
Huy Do committed
5378feb
T
[Windows] Disable constexpr std::mutex ctor so wheels load with msvcp140.dll older than 14.40 (#169429) (#196987)
Taras committed
2d02b07