strisunshinewentingwang/pytorch

Pull Requests Commits

remove unused THCBlas (#49725)

ngimel committed 3 years ago

04e04abd

Minor doc fix: change truncating to rounding in TF32 docs (#49625)

pbialecki committed 3 years ago

1451d847

Fix get_overlap_status for tensors without storage (#49638)

asuhan committed 3 years ago

21398fb6

Reland: Add base forward grad logic (#49734)

albanD committed 3 years ago

c23808d8

[onnxifi] Get rid of class member (#49380)

khabinov committed 3 years ago

eabe05ab

Make PyTorch partially cross-compilable for Apple M1 (#49701)

malfet committed 3 years ago

7b4a7661

[ROCm] add 4.0 to nightly builds (#49632)

jeffdaily committed 3 years ago

42b5601f

Complex backward for torch.sqrt (#49461)

anjali411 committed 3 years ago

4d9d03fe

[fix] inplace remainder/% (#49390)

kshitij12345 committed 3 years ago

2df249f0

NewModuleTest: Don't call both check_jacobian and gradcheck (#49566)

zou3519 committed 3 years ago

dfb7520c

[Gradient Compression] Warm-start of PowerSGD (#49451)

Yi Wang committed 3 years ago

c348faed

[PyTorch] Remove direct reference to native symbols in sparse related non-native codes (#49721)

iseeyuan committed 3 years ago

590e7168

[pt][ATen] Optimize bmm (#49506)

Hao Lu committed 3 years ago

d54cf2aa

[FX] Fix python code having spurious newlines from placeholders (#49720)

James Reed committed 3 years ago

11598da2

fx quant: fix types on _find_quants (#49616)

vkuzo committed 3 years ago

edce6b13

fx quant: add types to observed_module.py (#49607)

vkuzo committed 3 years ago

7c90b20f

fx quant: types for fusion_patterns.py (#49606)

vkuzo committed 3 years ago

9d5d1937

unbreak mypy torch/quantization (#49549)

vkuzo committed 3 years ago

ab2194f9

[TensorExpr] Move `SimpleIREval` implementation from .h to .cpp. (#49697)

Mikhail Zolotukhin committed 3 years ago

a5b27d7a

[TensorExpr] Change `LoopNest::vectorize` to accept `For*` instead of `Stmt*`. (#49696)

Mikhail Zolotukhin committed 3 years ago

e1f73ced

Revert D25607503: Add base forward grad logic

Walter Shen committed 3 years ago

f5178bf1

replacing THC_CLASS and THC_API with TORCH_CUDA_API (#49690)

janeyx99 committed 3 years ago

aa2782b9

Fix TCPStore type coercion (#49685)

H-Huang committed 3 years ago

7eb392d7

Use store based barrier only for certain store types. (#49694)

pritamdamania committed 3 years ago

1043ecf6

Move device guard from MultiTensorApply.cuh (#46664)

Iurii Zdebskyi committed 3 years ago

7e1356db

[jit][tracer] allow traced modules to return dicts with tuple values when strict=False (#49568)

bradleyhd committed 3 years ago

5b163e23

Do not use negative values in GCD computation. (#49379)

navahgar committed 3 years ago

46c9a0e6

Add base forward grad logic (#49097)

albanD committed 3 years ago

fdf02eff

Fix test_cuda_init_race skip rules (#49693)

malfet committed 3 years ago

befe3370

Enable product for bool tensor (#48637)

Kiyosora committed 3 years ago

983bfc79

Newer Older