strisunshinewentingwang/pytorch

Pull Requests Commits

Update over-cautious heuristic in AOTAutograd

Chillee committed 2 years ago

Verified 138c16fa

[ao][sparsity] Refactor and bugfix SparseQuantizedLayer tests (#81802)

HDCharles committed 2 years ago

8d693492

[ao][sparsity][fx] making fx qat and sparsity compose (#82204)

HDCharles committed 2 years ago

686a0941

optimize softmax backward and logsoftmax backward (#80114)

jiayisunx committed 2 years ago

15a284b0

Upgrade fbgemm in OSS PyTorch (#82676)

jianyuh committed 2 years ago

916a5651

Merge torchdim into functorch build (#82454)

zdevito committed 2 years ago

23b90044

[Re-land] [CUDA graphs] Clear autocast amp cache (#81896)

Aidyn-A committed 2 years ago

da0a3fe0

Fix nightly docs (#82672)

janeyx99 committed 2 years ago

0fd990ef

[MPS] Unary ops over empty tensors should be no-op (#82650)

malfet committed 2 years ago

420c5768

Update model_ops.yaml (#82444)

linbinyu committed 2 years ago

9e81c0c3

[JIT] refactor SchemaInfo to lazily initialize (#82596)

davidberard98 committed 2 years ago

149f0eb5

Revert "Revert "Add a lint rule for PYBIND11_DECLARE_HOLDER_TYPE (#82556)"" (#82600)

ezyang committed 2 years ago

428e0246

Revert "Revert "Add a lint rule for torch/csrc/util/pybind.h include (#82552)"" (#82599)

ezyang committed 2 years ago

df696608

Revert "[functorch] Fix linalg batch rules to error on non-matrix inputs (#82176)"

pytorchmergebot committed 2 years ago

31292599

Revert "Remove unecesary copy constructor (#82626)"

pytorchmergebot committed 2 years ago

dba287ed

[JIT] Refactor SchemaInfo training ops to account for rrelu_with_noise case (#82441)

goldenxuett committed 2 years ago

d326a55b

[MPS] Add dispatch stub code for MPS backend. (#82612)

kulinseth committed 2 years ago

19e3ea5e

reland "add a reinplacing FX pass (#80897)" (#82407)

bdhirsh committed 2 years ago

d362b8e9

[mergebot] Amend messages for land checks (#82649)

zengk95 committed 2 years ago

95c3e237

Revert #75195 (#82504)

zou3519 committed 2 years ago

8f86e361

[FSDP] Implement _param_fqns() to return all parameter FQNs for the FSDP module (#82595)

fegin committed 2 years ago

c05c3952

check params shape for mkldnn_convolution (#76526)

liukun committed 2 years ago

9f5f6ba6

[TorchTidy] Reland #81941 Add pattern to detect if bias is enabled in conv2d followed by batchnorm2d (#82421)

davidchencsl committed 2 years ago

ae2e303d

[tensorboard] Remove dependence on torch.onnx (#82628)

justinchuby committed 2 years ago

c7dde7da

Remove unecesary copy constructor (#82626)

Robert Serrano Kobylyansky committed 2 years ago

714669e2

Pin windows numpy (#82652)

janeyx99 committed 2 years ago

f1f231c4

Bumping nvidia docker version and using python 3.10 for cuda11.7 (#82472)

atalman committed 2 years ago

5aef0351

Replace SymIntTable with directly storing pointer in SymInt (#82468)

ezyang committed 2 years ago

b2bb2251

Attach ProxyTorchDispatchMode to ProxyTensor and use the mode in __torch_dispatch__ (#82549)

IvanYashchuk committed 2 years ago

a3316cb3

[functorch] Fix linalg batch rules to error on non-matrix inputs (#82176)

samdow committed 2 years ago

1dfcad84

Older