-
Notifications
You must be signed in to change notification settings - Fork 28.9k
Pull requests: pytorch/pytorch
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[Testcase Refactoring] Decouple test_benchmarking.py from CUDA-specific
ciflow/inductor
module: inductor
module: tests/graph
This label covers all test cases within the Graph sheet.
open source
topic: not user facing
topic category
#193783
opened Aug 17, 2026 by
pengyeqing
•
Draft
[CUDA] Validate MultiMarginLoss targets in backward
open source
release notes: nn
release notes category
#193782
opened Aug 17, 2026 by
HU-qingqing
Contributor
Loading…
Widen Vectorized<Half>::fmod to float on NEON
module: cpu
CPU specific problem (e.g., perf, algorithm)
open source
release notes: cpu (aarch64)
release notes category for aarch64, arm, etc.
topic: performance
topic category
#193780
opened Aug 17, 2026 by
krrishapatel
Loading…
Remove stranded affine-quantization observer block from torch.ao.quantization
ciflow/inductor
open source
release notes: export
release notes: quantization
release notes category
topic: bc breaking
topic category
#193779
opened Aug 17, 2026 by
egeboy35
Loading…
[Testcase Refactoring] Split Capability.distributed.fsdp into fsdp1/fsdp2, add sharded_tensor
open source
#193778
opened Aug 17, 2026 by
shichuanliujie9527
•
Draft
Fix torch.remainder giving a zero result the dividend's sign
module: cpu
CPU specific problem (e.g., perf, algorithm)
module: cuda
Related to torch.cuda, and CUDA support in general
module: mps
Related to Apple Metal Performance Shaders framework
open source
release notes: python_frontend
python frontend release notes category
topic: bug fixes
topic category
#193776
opened Aug 17, 2026 by
krrishapatel
Loading…
[Testcase Refactoring] Decouple test_best_config.py from CUDA-specific
ciflow/inductor
module: inductor
module: tests/graph
This label covers all test cases within the Graph sheet.
open source
topic: not user facing
topic category
#193775
opened Aug 17, 2026 by
pengyeqing
•
Draft
[RFC] GB300 Inductor Reduction Fix
ciflow/inductor
ciflow/torchtitan
Run TorchTitan integration tests
meta-exported
module: inductor
#193774
opened Aug 17, 2026 by
Roger-ZRZ
Loading…
Stabilize prod higher-order gradients
open source
#193773
opened Aug 17, 2026 by
HU-qingqing
Contributor
Loading…
Validate lambd in ASGD
open source
release notes: optim
#193772
opened Aug 17, 2026 by
shashvat-singham
Loading…
[Testcase Refactoring] Decouple test_binary_folding.py from CUDA-specific
ciflow/inductor
module: inductor
module: tests/graph
This label covers all test cases within the Graph sheet.
open source
topic: not user facing
topic category
#193771
opened Aug 17, 2026 by
pengyeqing
•
Draft
Add a PrivateUse1 from_blob hook for backend-specific TensorImpl/StorageImpl construction
open source
#193770
opened Aug 17, 2026 by
aeytkn
Loading…
[Testcase Refactoring] Decouple test_comm_analysis.py from CUDA-specific
ciflow/inductor
module: inductor
module: tests/graph
This label covers all test cases within the Graph sheet.
open source
topic: not user facing
topic category
#193769
opened Aug 17, 2026 by
pengyeqing
•
Draft
[inductor] AOTI: skip fallback for aten.sym_stride.int in lite mode
ciflow/inductor
ciflow/torchtitan
Run TorchTitan integration tests
meta-exported
module: inductor
release notes: inductor (aoti)
#193768
opened Aug 17, 2026 by
zoranzhao
Member
Loading…
[inductor] AOTI: handle SymInt in a Scalar arg for proxy-executor fallback kernels
ciflow/inductor
ciflow/torchtitan
Run TorchTitan integration tests
meta-exported
module: inductor
release notes: inductor (aoti)
#193767
opened Aug 17, 2026 by
zoranzhao
Member
Loading…
Compute floor_divide in float for Half and BFloat16
module: cpu
CPU specific problem (e.g., perf, algorithm)
module: cuda
Related to torch.cuda, and CUDA support in general
open source
release notes: python_frontend
python frontend release notes category
topic: bug fixes
topic category
#193766
opened Aug 17, 2026 by
krrishapatel
Loading…
[fake tensor] Cache _unsafe_view dispatch
ciflow/inductor
#193765
opened Aug 17, 2026 by
anijain2305
Contributor
Loading…
[dynamo] Reuse the stack trace when consecutive fx nodes share one
ciflow/inductor
ciflow/torchtitan
Run TorchTitan integration tests
module: dynamo
#193764
opened Aug 17, 2026 by
anijain2305
Contributor
Loading…
Fix fmod and remainder returning NaN on the vectorized CPU path
module: cpu
CPU specific problem (e.g., perf, algorithm)
open source
release notes: python_frontend
python frontend release notes category
topic: bug fixes
topic category
#193763
opened Aug 17, 2026 by
krrishapatel
Loading…
[Testcase Refactoring] Refactor test_hooks.py for test cases classification and device generalization
ciflow/inductor
module: dynamo
module: tests/graph
This label covers all test cases within the Graph sheet.
open source
topic: not user facing
topic category
#193762
opened Aug 17, 2026 by
littlexiaozhang
Contributor
•
Draft
[Testcase Refactoring] Decouple test_compile_subprocess.py from CUDA-specific
ciflow/inductor
module: inductor
module: tests/graph
This label covers all test cases within the Graph sheet.
open source
topic: not user facing
topic category
#193761
opened Aug 17, 2026 by
pengyeqing
•
Draft
[Testcase Refactoring] Decouple test_config.py from CUDA-specific
ciflow/inductor
module: inductor
module: tests/graph
This label covers all test cases within the Graph sheet.
open source
topic: not user facing
topic category
#193759
opened Aug 17, 2026 by
pengyeqing
•
Draft
Save mutated arguments after the kernel call in generated autograd kernels
ciflow/trunk
Trigger trunk jobs on your pull request
open source
release notes: nn
release notes category
#193758
opened Aug 17, 2026 by
yinjiew
Loading…
2 of 4 tasks
misc refactoring/code movement
ciflow/inductor
ciflow/torchtitan
Run TorchTitan integration tests
module: dynamo
#193750
opened Aug 17, 2026 by
liangel-02
Contributor
Loading…
Fix 7 stale CODEOWNERS paths, remove 6 dead rules
open source
topic: not user facing
topic category
#193745
opened Aug 17, 2026 by
DanielSwift1992
Loading…
Previous Next
ProTip!
Filter pull requests by the default branch with base:main.