strisunshinewentingwang/pytorch

Pull Requests Commits

fix torch.jit.tracing for at::lift

bdhirsh committed 2 years ago

8d3f51e2

[ONNX] Refactor to remove inline imports - attempt 2 (#77448)

justinchuby committed 2 years ago

563c2719

Pretty-print dataclasses (#76810)

francescocastelli committed 2 years ago

dca416b5

add native kernel for weight_norm on CPU

mingfeima committed 2 years ago

6fa20bdf

[FSDP] Add unittests to test distributed checkpointing

fegin committed 2 years ago

a0cc38ea

Add ccol_indices and row_indices methods.

pearu committed 2 years ago

88205886

[DataLoader] Adding ability to use dill to pass DataPipes in mutiprocessing

VitalyFedyunin committed 2 years ago

edffd595

[FSDP] Implement sharded_state_dict and load_sharded_state_dict

fegin committed 2 years ago

58c9d521

[PT-D][Sharding] Enable more ops needed in the transformer model training

fduwjj committed 2 years ago

a2cb94d2

Simplify IfThenElse and CompareSelect within for-loop (#76793)

EikanWang committed 2 years ago

e5a5cd14

[mobile] add buck build for mobile targets (#76480)

linbinyu committed 2 years ago

168dc70f

Discover and check operator variants

amjames committed 2 years ago

e5a752a6

Connect Tensor.__ipow__ to pow_ method

amjames committed 2 years ago

39d3a7ff

Fix `mean` bug for integral tensors. (#76584)

ysiraichi committed 2 years ago

459e10f4

[NVFuser] fix force-disable flag

davidberard98 committed 2 years ago

e175065c

[fix] logsumexp: type promotion (integral to float) (#77480)

kshitij12345 committed 2 years ago

77ac82c7

Revert "add BFloat16 support for BatchNorm on CPU"

pytorchmergebot committed 2 years ago

93a96922

MPS: fixes (#77462)

kulinseth committed 2 years ago

978304fc

add BFloat16 support for BatchNorm on CPU

mingfeima committed 2 years ago

7c8911ca

[Reland] load_state_dict post hook (#77392)

rohan-varma committed 2 years ago

a275491c

improve softmax lastdim performance on bfloat16 by adding more fusion

mingfeima committed 2 years ago

2166fc55

improve group_norm channels last performance on CPU

mingfeima committed 2 years ago

59b56ba7

Register more decomps for meta.

ezyang committed 2 years ago

2f602abf

Revert "[LT] Codegen ReuseNode for supported ops"

pytorchmergebot committed 2 years ago

3cade9d4

Revert "[chalf] enable testing for multiple ops (#77405)"

pytorchmergebot committed 2 years ago

d0ce7921

Make s3_init linter adapter work in abscense of git (#77466)

malfet committed 2 years ago

ab1b4aa2

[GHF] Skip internal checks for bot comments (#77465)

malfet committed 2 years ago

e9de3a69

Revert "[quant][core][gpu][feature] Implemented quantized cuda gelu"

pytorchmergebot committed 2 years ago

981719fe

torch/distributed: move quantization ops to libtorch instead of libtorch_python (#77444)

d4l3k committed 2 years ago

ed51b930

[NVFuser] don't decompose conv2d if we don't have shape info

davidberard98 committed 2 years ago

36f7a6cc

Older