pytorch/benchmark

Pull Requests Commits

Guard OrderedDict linked order natively (#203342)

generatedunixname499836121 committed 3 hours ago

83279651

Remove deprecated public TorchScript APIs + tests

janeyx99 committed 1 day ago

6faac59b

Revert D129970553: Retire deprecated public TorchScript APIs; retain internal implementations only

janeyx99 committed 3 days ago

af3ab024

Retire deprecated public TorchScript APIs; retain internal implementations only

janeyx99 committed 3 days ago

584ec7e9

Add a user-kernel HOP registry and register Triton with it

zoranzhao committed 4 days ago

c960e717

A Dynamo artifact imports the modules that register its custom ops (#202258)

bobrenjc93 committed 10 days ago

4f46dd94

Route internal callables through builders (#199832)

williamwen42 committed 10 days ago

1c083b8a

Remove Python 3.10 support code (#199417)

generatedunixname499836121 committed 13 days ago

25ad5a60

Classify function tests by hardware (#193120)

generatedunixname499836121 committed 18 days ago

70d61077

fbcode/pytorch/benchmark/torchbenchmark/models/yolov3

generatedunixname2070776263773646 committed 20 days ago

f4ef5581

Seed `___unnamed_scope_*` globals when installing a package (#200823)

bobrenjc93 committed 23 days ago

9dd1f261

Disable nested graph breaks across iterator host frames (#199218)

williamwen42 committed 25 days ago

61227cc1

Fix B001: Replace bare except with except Exception

generatedunixname1634730417729618 committed 27 days ago

f8efcd15

Decouple benchmark device handling from CUDA and XPU (#192472)

generatedunixname499836121 committed 27 days ago

70acc828

Call bound builtin methods by descriptor, not by name (#199881)

generatedunixname499836121 committed 27 days ago

50a2b436

Run graphs that inlined a TorchFunctionMode with torch function disabled (#196056)

ezyang committed 34 days ago

ab5e0c9b

Fix is_pybind11_enum_member for Enum classes with a `name` member (#199043)

generatedunixname499836121 committed 35 days ago

f331b987

Run the PR test on OSDC runners (#2711)

huydhn committed 38 days ago

284984a8

Turn off Dynamo's LRU cache reordering on free-threaded builds

alexmalyshev committed 39 days ago

88e8837b

Build the nightly docker with remote BuildKit on OSDC, and move it to CUDA 13.0 (#2712)

huydhn committed 39 days ago

7bd8375c

Fix NumPy ndarray operators under default device contexts (#185462)

jansel committed 40 days ago

93c8f75b

Show carets in graph break user stacks (#185466)

jansel committed 40 days ago

0e5a95be

Remove `_base_vt` (#193679)

georgehong committed 40 days ago

70a2f537

Run the docker cleanup on a GitHub runner (#2710)

huydhn committed 41 days ago

4606439e

Add Qwen3.5-0.8B to Hugging Face benchmarks (#198261)

kevzhuo committed 41 days ago

cc2b9a40

Add support for list_reverseiterator (#197454)

generatedunixname499836121 committed 42 days ago

72b45587

Increase small tensor tolerance for functorch_dp_cifar10 (#198210)

williamwen42 committed 45 days ago

69e60455

Improve speed of checked device index (#195951)

generatedunixname499836121 committed 45 days ago

d205a01d

Bitwise float compare for guard snapshots and subgraph reuse (#192608)

ezyang committed 54 days ago

d5a9b67d

Seed the recorded import aliases and builtins-dict key into the guard scope on load (#196817)

bobrenjc93 committed 55 days ago

bd8759d5

Older