strisunshinewentingwang/pytorch

Pull Requests Commits

[PyTorch] Add native fast path for transformer encoder inference

swolchok committed 2 years ago

b369b89f

Rename tools/codegen to torchgen (#76275)

ezyang committed 2 years ago

36420b5e

[ONNX] Support restricted quantized range for activation.

BowenBao committed 2 years ago

8d31706b

[ONNX] Support per channel quantization

BowenBao committed 2 years ago

cada2cd3

[Distributed tests] Add skip for odd world_size condition

jaglinux committed 2 years ago

6ca8272d

Improving typing and typing-related performance in rendezvous.py

ishaan-mehta committed 2 years ago

d65ab9a6

Fixed typo in documentation of `LazyModuleMixin`

sugatoray committed 2 years ago

f980c3c1

Fix minor typos

meyavuz committed 2 years ago

89370e00

bypassed cublasLtMatMul bug after slicing (#76205)

xiaosun86 committed 2 years ago

a5ffdaf0

[Model Averaging] Make the error message more informative in hierarchical_model_averager.py

wayi1 committed 2 years ago

e9058039

Revert "stft: remove non-center overload and python functional wrapper"

pytorchmergebot committed 2 years ago

77f23d64

Emit ATen ops when symbolics raise + minor fixes

Thiago Crepaldi committed 2 years ago

90d31cb3

[nnc] Strides to Tensor (#72962)

IvanKobzarev committed 2 years ago

93906092

Remove incorrect xfail

peterbell10 committed 2 years ago

1a7e43be

Enable simple indexing into CSR tensor, add torch.select for CSR

IvanYashchuk committed 2 years ago

407e8eba

[Pytorch][3/4 Static dispatch] Move static dispatch logic to Operators.cpp (#76058)

priyaramani committed 2 years ago

5622c8b4

Make LTC codegen customizable enough for XLA migration (#76180)

wconstab committed 2 years ago

9d8ff02e

Remove cuda 11.5 builds since we have 11.6

atalman committed 2 years ago

31a6e6ca

stft: remove non-center overload and python functional wrapper

peterbell10 committed 2 years ago

6b7d89c4

suppors dynmic_cast, added uts

SherlockNoMad committed 2 years ago

cd00e108

Remove F.pad python implementation

peterbell10 committed 2 years ago

cb37e7a0

[fx] Don't use __module__ to test if a function is bound from C++

peterbell10 committed 2 years ago

7a80fc2c

[ONNX] Add typing annotations to onnx symbolic gelu

justinchuby committed 2 years ago

2f2158ae

Revert "Dnt CSE across context managers"

pytorchmergebot committed 2 years ago

a71fabab

[JIT] Reuse traced fn for jit opinfos

davidberard98 committed 2 years ago

1324410f

Added directory check before saving in C++ API

vitrioil committed 2 years ago

75574076

fix UndefinedBehaviorSanitizer

anjali411 committed 2 years ago

79891abf

[ROCm] enable fsdp tests

jeffdaily committed 2 years ago

44bbb247

Make GenLazyNativeFuncDefinition abstract and extensible (#75343)

wconstab committed 2 years ago

88b61d13

Fix to support no-batch-dim inputs in ConvTransposeNd._output_padding

jbschlosser committed 2 years ago

041e6e75

Newer Older