strisunshinewentingwang/pytorch

Pull Requests Commits

equality assertions (#102256)

avikchaudhuri committed 1 year ago

87510022

[PT2][Quant][BE] Apply formatting to test_quantize_pt2e (#102275)

kimishpatel committed 1 year ago

9b5e4c30

Document faster builds for C++ changes (#102316)

malfet committed 1 year ago

efd774a2

Bump requests from 2.30.0 to 2.31.0 in /tools/build/bazel (#102059)

dependabot[bot] committed 1 year ago

Verified c05a3173

Revert "add additional stream priority for cuda streams (#101956)"

pytorchmergebot committed 1 year ago

6c9b94dc

[MTPG] Enable for some tests in test_fsdp_misc (#102043)

rohan-varma committed 1 year ago

3dfa755a

Add cpp.max_horizontal_fusion_size to control the granularity of horizontal fusion (#99828)

EikanWang committed 1 year ago

ce41faa2

[vision hash update] update the pinned vision hash (#102318)

pytorchupdatebot committed 1 year ago

e1dc793e

[ONNX] Support aten::scatter_reduce (#102048)

titaiwangms committed 1 year ago

fb468b67

Increase mem eff backward performance (#101847)

drisspg committed 1 year ago

ef13fde2

Invoke the bf16 load w/o #elements to bypass the temporary buffer allocation from the performance perspective. (#99822)

EikanWang committed 1 year ago

6f464e0c

Add fast path for BF16 kernel if all the operations within the kernel support bf16 (#99814)

EikanWang committed 1 year ago

c3550d83

Remove inplace buffers when original and mutation are both removed (#102289)

Natalia Gimelshein committed 1 year ago

68816e4f

[OpInfo] Add multi_head_attention_forward (#100153)

qqaatw committed 1 year ago

0db704d2

Revert "Disallow _foreach_utils.py, but allow it to be inlined (#102221)"

pytorchmergebot committed 1 year ago

8aa48315

[CUDA][DLPack] Try ~~bumping sleep interval~~ running on explicit side-stream for Windows `dlpack` test (#102283)

eqy committed 1 year ago

54f38381

Integrating new API usage metadata logger (#101762)

atannous committed 1 year ago

b469ed72

Make test_inductor_collectives use self.assert* (#102274)

wconstab committed 1 year ago

ae5606bb

simplify BinaryDivFloorKernel.cu code (#102168)

XiaobingSuper committed 1 year ago

b628eb52

Disallow _foreach_utils.py, but allow it to be inlined (#102221)

mlazos committed 1 year ago

552299c4

[MPS] Allow saved models to be loaded directly to MPS through torch.jit.load (#102204)

DenisVieriu97 committed 1 year ago

de7ec2dd

[inductor] Support precomputed_sizes in CppWrapperCodeGen (#102083)

desertfire committed 1 year ago

836798e0

[ONNX] Bump ORT version to 1.15.0 (#102248)

titaiwangms committed 1 year ago

053dff11

fix benchmarks/dynamo/runner.py (#102311)

williamwen42 committed 1 year ago

3c773107

DOC Fixes unpacking issue in dynamo explain docs (#101761)

thomasjpfan committed 1 year ago

0d17bd5f

fix functorch/test_ops.py test_vjp flash attention unexpected success (#102131)

Fuzzkatt committed 1 year ago

5b01c8dc

[ez] add `docs/source/compile/generated/` to .gitignore (#101094)

dingran committed 1 year ago

184d4f1b

Foreach kernel codegen in inductor (#99975)

mlazos committed 1 year ago

80f72648

[CUDAGraph Trees] Fix empty storages handling (#102273)

eellison committed 1 year ago

1f80b972

[dynamo] fix module buffers call (#102251)

wanchaol committed 1 year ago

c1db2350

Newer Older