strisunshinewentingwang/pytorch

Pull Requests Commits

Add forward AD support for logsumexp, log_softmax, softmax, nll_loss, and cross_entropy (#73741)

soulitzer committed 2 years ago

de73f9a5

Update lazy_bench.py (#73889)

wconstab committed 2 years ago

Verified 1d12b2ba

Implement TensorList and demonstrate it via Stack op (#73758)

wconstab committed 2 years ago

Verified 353aece7

[RELAND] [cuDNN] Add a new optimized cuDNN RNN algorithm for small RNN hidden_size (#73211)

mcarilli committed 2 years ago

943085fe

Update __torch_dispatch__ to return op overload instead of the opoverload packet function (#72673)

anjali411 committed 2 years ago

086645ad

[JIT] make RegisterCudaFuseGraph use TORCH_API instead of C10_EXPORT (#73742)

davidberard98 committed 2 years ago

a3d099ea

[ONNX] Fix repeat interleave when repeats and dim is 1

BowenBao committed 2 years ago

6c26bf0e

[FSDP][BE] Change assert to assertEqual (#73787)

awgu committed 2 years ago

15a799e3

[Easy][FSDP] Fix warning render (#73786)

awgu committed 2 years ago

89b16f1a

[FSDP] Add grad accumulation without `no_sync()` (#73535)

awgu committed 2 years ago

4a06b8d3

[Static Runtime] Add out variant wrapper for aten::ones (#73851)

d1jang committed 2 years ago

71961d37

ci: Migrate windows conda to GHA

seemethere committed 2 years ago

2acf9c74

ci: Enable long paths for windows

seemethere committed 2 years ago

07410207

CMake option for using static MKL libraries

ashari4 committed 2 years ago

7ed73b28

[Quant][fx] Add lowering for functional conv (#73708)

andrewor14 committed 2 years ago

f3c6e8f7

[Quant][fx] Add lowering for Linear-Bn1d in QAT mode (#73509)

andrewor14 committed 2 years ago

cedce3be

[GHA] Migrate win/linux binary-smoke workflows from CircleCI

malfet committed 2 years ago

9929a9fc

HaoZeke committed 2 years ago

979a78f8

[Metal] Use MPSNNNeuronDescriptor for Mac Catalyst Target (#73718)

xta0 committed 2 years ago

1d497114

[Metal] Remove unused shader code (#73636)

xta0 committed 2 years ago

e118b1df

Revert "[GHA] Migrate win/linux binary-smoke workflows from CircleCI"

pytorchmergebot committed 2 years ago

ea698c14

Revert D34230284: [pytorch][PR] [WIP][Kineto] Manual Submodule Update

malfet committed 2 years ago

e7730eae

[GHA] Migrate win/linux binary-smoke workflows from CircleCI

malfet committed 2 years ago

486bd9f3

[WIP][Kineto] Manual Submodule Update (#73090)

aaronenyeshi committed 2 years ago

c088c8fe

Automated submodule update: FBGEMM (#73814)

facebook-github-bot committed 2 years ago

d047e475

[XNNPACK] Expose available() to xnnpack namespace (#73617)

digantdesai committed 2 years ago

4b8106f9

[Metal] Raise the minimum support version of iOS to 11.0 (#73635)

xta0 committed 2 years ago

5dbec7c0

CUDA Kernels: Use per-operator headers (1/4) (#71212)

peterbell10 committed 2 years ago

723ba4e3

CPU Kernel: Use per-operator headers (#71137)

peterbell10 committed 2 years ago

5816a95c

ns for fx: add ability for fp16 model to shadow fp32 model (#73785)

vkuzo committed 2 years ago

6f2dad24

Newer Older