strisunshinewentingwang/pytorch

Pull Requests Commits

Remove inferred from tensor type ctors (#48263)

Elias Ellison committed 3 years ago

0b9d5e65

[TensorExpr Fuser] Add support for nodes which have tensor constant inputs (#47814)

Elias Ellison committed 3 years ago

71ddc0ba

[NNC] Compute Tensor Output Properties in ininitialization (#47813)

Elias Ellison committed 3 years ago

413caa7f

[TensorExpr] Cache use of fallback in kernel invocation (#47812)

Elias Ellison committed 3 years ago

0e666a9f

Dont use symbolic shapes check (#47810)

Elias Ellison committed 3 years ago

70853c50

make duplicate def() calls an error in the dispatcher (#48098)

bdhirsh committed 3 years ago

18c03b9f

[Binary Push] Update the awscli installation, use conda install rather than brew install (#49175)

husthyc committed 3 years ago

2519348f

[iOS] Bump up the cocoapods version (#49176)

husthyc committed 3 years ago

edbf9263

[vmap] implement batching rule for fill_ and zero_ (#48516)

RockingJavaBean committed 3 years ago

909a9060

Check CUDA kernel launches (/fbcode/caffe2/) (#49145)

Yixin Bao committed 3 years ago

840e71f4

Use new FFT operators in stft (#47601)

peterbell10 committed 3 years ago

524adfbf

Add missing complex support for torch.norm and torch.linalg.norm (#48284)

kurtamohler committed 3 years ago

54f0556e

add additional interpolation modes for torch.quantile (#48711)

RockingJavaBean committed 3 years ago

25a8397b

Refactor cudnn convolution (#49109)

zasdfgbnm committed 3 years ago

45473ffe

Allow ROCm CI to use non-default stream. (#48424)

jeffdaily committed 3 years ago

d5c4a80c

Revert D25441716: [te] Add BitCast to the IR

bwasti committed 3 years ago

195b92bf

[te] Add BitCast to the IR

bwasti committed 3 years ago

33841454

make AT_FFTW_ENABLED available to fb internal

Rong Rong committed 3 years ago

21c04b44

fix some comments in accelerator_partitioner.py (#49104)

scottxu0730 committed 3 years ago

33bc7918

Decouple direct access to native::scalar_tensor from TensorIndexing.h (#48761)

iseeyuan committed 3 years ago

c7b8f3e2

Revert D25433268: [PyTorch Mobile] Preserve bundled input related methods when calling optimize_for_mobile

lw committed 3 years ago

2255e68d

Cache the DataPtrs in CUDAFuture (#48788)

lw committed 3 years ago

b5a7e250

Split out reusable CUDAFuture from FutureNCCL (#48506)

lw committed 3 years ago

030fa6cf

Merge common parts of FutureNCCL into at::ivalue::Future (#48505)

lw committed 3 years ago

4c425e8d

Split FutureNCCL's CUDA-specific parts from generic future logic (#48504)

lw committed 3 years ago

9078088e

Support wider range of types in FutureNCCL (#48502)

lw committed 3 years ago

a6778989

Don't store device indices separately on FutureNCCL (#48501)

lw committed 3 years ago

9fe3ac36

Add multi-GPU support to FutureNCCL (#48500)

lw committed 3 years ago

e294c2d8

Fix FutureNCCL not recording dataptrs with caching alloc in wait() (#48563)

lw committed 3 years ago

91ad3ed8

Fix FutureNCCL's completed() disagreeing with wait() (#48503)

lw committed 3 years ago

003c30ba

Newer Older