strisunshinewentingwang/pytorch

Pull Requests Commits

[HT] Clear the device placement tag for the auto gen sum so that we could break the component for FC sharing the same input (#42219)

Xing Wang committed 4 years ago

27b03d62

fix typo in GradScaler docstring (#42236)

mcarilli committed 4 years ago

7cdf786a

grad detach_ only when it has grad_fn in zero_grad call (#41283)

zhaojuanmao committed 4 years ago

79cfd859

Creates spectral ops test suite (#42157)

Mike Ruberry committed 4 years ago

4b6e5f42

Improved coverage for unboxed->boxed kernel wrappers (#38999)

Basil Hosmer committed 4 years ago

029007c8

[Caffe2] Fix spatial_bn bug for computing running_var on CPU or on CUDA without CuDNN (#42151)

xiaomengy committed 4 years ago

60f51542

Environment variable for controlling type verbosity in debug output (#41906)

bertmaher committed 4 years ago

91546a4b

Operator-level Benchmark Test for Per Tensor and Per Channel Fake Quantization (#41974)

Paul Shao committed 4 years ago

01b794f1

add tests to BinaryOpsKernel -- max/min kernel (#42198)

scintiller committed 4 years ago

48acdfd5

Extending Learnable Fake Quantize module to support gradient scaling and factory (partial) construction (#41969)

Paul Shao committed 4 years ago

38278122

[JIT] Dont include view ops in autodiff graphs (#42027)

Elias Ellison committed 4 years ago

0a64f991

Fix type annotation for DistributedDataParallel (#42231)

Zhijian Liu committed 4 years ago

b45b82b0

Automated submodule update: FBGEMM (#42205)

facebook-github-bot committed 4 years ago

c8e15842

Revert D22790718: [pytorch][PR] Enables torch.full bool and integer type inference

albanD committed 4 years ago

46097048

implement numpy-like functionality isposinf, isneginf (#41588)

RockingJavaBean committed 4 years ago

90074bbf

[pt] Add incude_last_offset option to EmbeddingBag mean and max (#42215)

jianyuh committed 4 years ago

1c5c289b

Enables torch.full bool and integer type inference (#41912)

Mike Ruberry committed 4 years ago

6b3f3356

DOC: fail to build if there are warnings (#41335)

mattip committed 4 years ago

8c653e05

refactor save_data as non member function (#42045)

ann-ss committed 4 years ago

4b108ca7

Remove dead named_tensors_unsupported_error definitions. (#42171)

ezyang committed 4 years ago

8fc5adc8

Fix flaky NCCL error handling tests. (#42149)

pritamdamania committed 4 years ago

8deb4fe8

Add appropriate error messages for ProcessGroupNCCLTest (#42143)

osalpekar committed 4 years ago

b6a9f427

Fixed Skipping Logic in ProcessGroupNCCLErrors tests (#42192)

osalpekar committed 4 years ago

e4c3f526

Add a flag to enforce fp32 to fp16 conversion for all inputs of the onnxifi net. (#39931)

ipiszy committed 4 years ago

b2ef7fa3

[Shape Inference] Fix InferFC

ChunliF committed 4 years ago

8a644f0c

[quant][graphmode] Support stack (#42187)

jerryzh168 committed 4 years ago

30eacb5f

Stop building PyTorch for VS2017 (#42144)

malfet committed 4 years ago

deac621a

Dequant => Swish => Quant Test case. (#41976)

venkatacrc committed 4 years ago

3c084fd3

Use Python3.7 when running OSX builds/tests (#42191)

malfet committed 4 years ago

e2344db8

make FusionCallback refer to specified GraphFuser context (#41560)

wconstab committed 4 years ago

4c7fb8c2

Newer Older