strisunshinewentingwang/pytorch

Pull Requests Commits

nikithamalgifb committed 3 years ago

bea152bb

model_dump tool for model inspection (#56868)

dreiss committed 3 years ago

89377e3e

change ddpLoggingData struct to map or dict (#56641)

zhaojuanmao committed 3 years ago

1e77ba36

[profiler] Support for trace metadata (#56575)

Ilia Cherniavskii committed 3 years ago

3115728c

Update floor_divide behavior in line with NumPy 1.20 (#56893)

peterbell10 committed 3 years ago

5536cda1

[profiler] Add cuda synchronization point (ci-all) (#57036)

Ilia Cherniavskii committed 3 years ago

77721ee3

[iOS GPU][Kernel] Implement channel split in Metal shaders (#56074)

xta0 committed 3 years ago

8134806e

Torchelastic: add support for the new error file format (#57084)

aivanou committed 3 years ago

0df57401

[caffe2] fix bug when weight_decay is used with fused rowwise + SLWS grad (#57090)

jspark1105 committed 3 years ago

882e2736

Add missing vec256<>::isnan() for VSX float and double vectors (#56658)

malfet committed 3 years ago

51e6ebb5

[PyTorch][Edge] Add binarires for unittests (#57039)

cccclai committed 3 years ago

c91ea7d4

[FX] fix normalization issues with lists of tensors (#57004)

Chillee committed 3 years ago

786b0a80

Fix clang-tidy for native CPU ops (#57037)

malfet committed 3 years ago

1c0617bb

[ARM] Do not use depthwise3x3 conv in grad mode (#56889)

malfet committed 3 years ago

808850b6

To fix inconsistency of digamma with SciPy (#56689)

iramazanli committed 3 years ago

6e826cac

[aten][simple] Optimize atrepeat (#56994)

Hao Lu committed 3 years ago

0319b64e

Remove sync for randperm on small tensors. (#54113)

zasdfgbnm committed 3 years ago

e8c26874

ns for fx: additional bugfix for user defined functions (#57028)

vkuzo committed 3 years ago

9fe2673d

ns for fx: allow comparing int8 to int8 for functionals (#57027)

vkuzo committed 3 years ago

da2cef6a

ns for fx: add option to skip matching classes and functions (#57026)

vkuzo committed 3 years ago

a359cfac

ns for fx: support binary ops when adding unshadowed loggers for inputs (#57025)

vkuzo committed 3 years ago

e8a5490c

ns for fx: bug fix for shadowing fp16 emulation patterns (#57024)

vkuzo committed 3 years ago

ddedeab6

ns for fx: add fp16 function shadowing (#57023)

vkuzo committed 3 years ago

2acc19ec

ns for fx: allow user functions in shadowing (#57022)

vkuzo committed 3 years ago

782a0a14

ns for fx: move node I/O dtype mapping to be local instead of global (#57021)

vkuzo committed 3 years ago

c4bec76b

Output tensor specified via out= must be on the same device as inputs for dot & vdot (#56334)

imaginary-person committed 3 years ago

c3073791

Revert D27765618: Initial support for sparse complex tensors constructors for CPU/CUDA

Mike Ruberry committed 3 years ago

7bcce2ac

fix #56822 (#56967)

eqy committed 3 years ago

fa57191b

Eliminate global usage of torch.set_default_dtype in sparse test (#56393)

aocsa committed 3 years ago

0d41122e

Modernize test-suite in sparse tensor CSR (#56392)

aocsa committed 3 years ago

18c89a90

Older