strisunshinewentingwang/pytorch

Pull Requests Commits

add simd horizantal reduce to improve log_softmax and softmax performance on CPU

mingfeima committed 2 years ago

dcc255d1

Make tracer be able to trace different forward functions (#77109)

sijiac committed 2 years ago

bf4b6d0d

Add new PrivateUse1 DeviceType for non-public devices (#77208)

yuhc committed 2 years ago

abb6fab0

Retry "[JIT] parse prim::Constant[value=annotate()] and prim::Constant[value={0}]"

davidberard98 committed 2 years ago

fa44e165

Index reduction CUDA support

mikaylagawarecki committed 2 years ago

1141b45e

fix StridesPolicy logic for FunctionalTensorWrapper

bdhirsh committed 2 years ago

5762c7b2

[BE] Fix shadowed variable warning in `c10::TensorType::contiguousStridesOf` (#77387)

malfet committed 2 years ago

9fcf75e5

[primTorch] reshape and view (#77220)

Mike Ruberry committed 2 years ago

64c6a89b

fix torchvhsion failed case test_classification_model on slow_conv2d

mingfeima committed 2 years ago

2b7943c4

[complex32] sum, prod : cuda only (disable jiterator reduction on windows) (#77192)

kshitij12345 committed 2 years ago

39bd37f3

Add trace and log_sigmoid_forward decomps (#77329)

Chillee committed 2 years ago

8626f765

[ONNX] Refactor to remove inline imports (#77142)

justinchuby committed 2 years ago

c08b8f09

make scalar clamp overloads propagate nan (#77371)

Natalia Gimelshein committed 2 years ago

5c7d916c

[JIT] Support for ParameterDict getattr

davidberard98 committed 2 years ago

09255977

[ONNX] Add quantization support for maxpool (#77393)

justinchuby committed 2 years ago

1dd73364

[ONNX] Relax node constraint for onnx shape inference (#77379)

BowenBao committed 2 years ago

a812c4cd

mul(dense, csr), mul(csr, dense) via sparse_mask_csr (#77177)

cpuhrsch committed 2 years ago

b2507592

[ONNX] Fix type comparison in utils._need_symbolic_context (#77365)

justinchuby committed 2 years ago

78d37981

ci: Add pr number to job name for trymerge

seemethere committed 2 years ago

65f71c0c

Make it possible to register decompositions to Meta key

ezyang committed 2 years ago

d5ed73ba

exposing more CUDA driver API (#77296)

jjsjann123 committed 2 years ago

1a919756

ci: Switch lint to use full custom runners

seemethere committed 2 years ago

69347dc5

ci: switch trymerge to custom runner

seemethere committed 2 years ago

b1214bab

ci: Remove unused azure_pipelines ci infra

seemethere committed 2 years ago

86e923be

Remove unnecessary ifdef, fixes fbcode build

ezyang committed 2 years ago

4ef64079

Revert "Load state dict post hook"

pytorchmergebot committed 2 years ago

d92b0a51

ci: Set create_release.yml to run on nightly

seemethere committed 2 years ago

3e92cae0

ci: Cleanup create_release.yml workflow

seemethere committed 2 years ago

35af5f3c

reenable filtered op tests (#77330)

Natalia Gimelshein committed 2 years ago

14f84a87

Support operator list in LightWeightDispatch

priyaramani committed 2 years ago

81e2708d

Newer Older