-
Notifications
You must be signed in to change notification settings - Fork 26.9k
Pull requests: pytorch/pytorch
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[MPS] Ensure pool2d inputs are contiguous in target memory format
release notes: mps
Release notes category
Work around eager-mode issues with logical-xor op
fb-exported
meta-exported
topic: not user facing
topic category
#175318
opened Feb 19, 2026 by
kasperl
Loading…
[DTensor] Use custom op for exact Replicate -> Partial(sum)
ciflow/inductor
release notes: distributed (dtensor)
release notes category
#175317
opened Feb 19, 2026 by
zpcore
Loading…
fix(elastic): respect USE_LIBUV on Windows so torchrun works without libuv
open source
release notes: distributed (c10d)
release notes category
#175316
opened Feb 19, 2026 by
bugparty
Loading…
[CPUBLAS] Fix UB: use vector::resize() instead of reserve() before operator[] access
open source
#175315
opened Feb 19, 2026 by
mulatta
Loading…
1 of 4 tasks
[DTensor] PQ search: fast path, cost cache, and reachability tests
ciflow/inductor
release notes: distributed (dtensor)
release notes category
#175313
opened Feb 19, 2026 by
wconstab
Loading…
WIP first batch of claude fixes, include PQentry and cost computation optimization
ciflow/inductor
release notes: distributed (dtensor)
release notes category
#175312
opened Feb 19, 2026 by
wconstab
Loading…
[inductor] Skip addcdiv decomposition for AdamW bitwise precision
ciflow/inductor
module: dynamo
module: inductor
#175310
opened Feb 19, 2026 by
mlazos
Loading…
[inductor] Skip addcmul decomposition to enable FMA lowering
ciflow/inductor
module: inductor
#175309
opened Feb 19, 2026 by
mlazos
Loading…
[Inductor] Allow tiling reduction output dimensions
ciflow/inductor
module: inductor
topic: not user facing
topic category
#175308
opened Feb 19, 2026 by
blaine-rister
•
Draft
Remove hardcoded CUDA version from backend configuration in test
ciflow/inductor
module: inductor
open source
topic: not user facing
topic category
#175307
opened Feb 19, 2026 by
crcrpar
Loading…
Use
Sm100CollectiveEpilogue on SM100
ciflow/inductor
module: inductor
open source
#175305
opened Feb 19, 2026 by
crcrpar
Loading…
[ROCM] Refactor BFloat16 implementation for native usage of BF16 to float conversion.
module: rocm
AMD GPU support for Pytorch
open source
#175303
opened Feb 19, 2026 by
anatoliylitv
•
Draft
[DTensor][compile] replay view sequence on subclasses
ciflow/inductor
#175301
opened Feb 19, 2026 by
pianpwk
Loading…
[CI] Move CUDA 12.8 GPU tests from per-commit trunk to periodic
open source
topic: not user facing
topic category
#175300
opened Feb 19, 2026 by
pytorchbot
Loading…
[benchmark] Skip pytorch_CycleGAN_and_pix2pix from inductor benchmarks
ciflow/inductor
module: dynamo
open source
topic: not user facing
topic category
#175299
opened Feb 19, 2026 by
pytorchbot
Loading…
[stack trace] ignore CrossRefMode in stack trace
ciflow/inductor
release notes: fx
release notes category
#175298
opened Feb 19, 2026 by
yushangdi
Loading…
[fx] Fix KeyError for unbacked_bindings in slice/cat operations
open source
release notes: fx
release notes category
#175297
opened Feb 19, 2026 by
Chevi-Koren
•
Draft
[vision hash update] update the pinned vision hash
ci-no-td
Do not run TD on this PR
ciflow/inductor
ciflow/trunk
Trigger trunk jobs on your pull request
open source
topic: not user facing
topic category
#175296
opened Feb 19, 2026 by
pytorchupdatebot
Loading…
[BE] Clone build venv when creating worktrees
topic: not user facing
topic category
#175295
opened Feb 19, 2026 by
malfet
Loading…
Update triaging rules
topic: not user facing
topic category
#175291
opened Feb 18, 2026 by
malfet
Loading…
Add cudagraph_exclude_sym_shape and cudagraph_include_sym_shape APIs
ciflow/inductor
module: dynamo
module: inductor
#175288
opened Feb 18, 2026 by
eellison
Loading…
[DTensor] view_groups flatten-split fallback
ciflow/inductor
release notes: distributed (dtensor)
release notes category
#175287
opened Feb 18, 2026 by
pianpwk
Loading…
Previous Next
ProTip!
Add no:assignee to see everything that’s not assigned.