strisunshinewentingwang/pytorch

Pull Requests Commits

srossross committed 2 years ago

ea101223

srossross committed 2 years ago

97767239

update test and gen

srossross committed 2 years ago

6ec45a96

move function to PlumbingHelper

srossross committed 2 years ago

29f03424

srossross committed 2 years ago

c837512f

fix test add _linalg_check_errors case

srossross committed 2 years ago

a2715e19

Improve from review

srossross committed 2 years ago

a35477a9

comments from review

srossross committed 2 years ago

26c167c4

local lintrunner did not catch this?

srossross committed 2 years ago

aa1582bb

srossross committed 2 years ago

d294370e

changed another assert to check

srossross committed 2 years ago

a3478491

update gen code and added test

srossross committed 2 years ago

12ddd882

[Pytorch][Vulkan] Use specalized shader for 3x3 depthwise conv (#89953)

kimishpatel committed 2 years ago

55b10e6b

[Pytorch][Vulkan] Templatize depth wise convolution and specialize for 3x3 and

kimishpatel committed 2 years ago

a17765a1

[Pytorch][Vulkan] shader codegen use ordered dictionary (#89951)

kimishpatel committed 2 years ago

bd456fb5

[Pytorch][vulkan] Simplify depthwise conv to remove bounds compute (#89950)

kimishpatel committed 2 years ago

cb68dcbd

[Vulkan] output benchmark numbers for aibench parsing (#89949)

kimishpatel committed 2 years ago

876b7024

[pytorch][vulkan] realistic benchmark size for depthwise (#89948)

kimishpatel committed 2 years ago

841eba63

[Caffe2] Fix the assert message (#89816)

atuljangra committed 2 years ago

564905c8

Fix XLA dynamo CI (#90229)

JackCaoG committed 2 years ago

2ea32f41

Add sharding strategy to fully_shard (#90192)

mrshenli committed 2 years ago

5d6aa99c

Add a repro for fully_shard _unshard error (#90190)

mrshenli committed 2 years ago

e4670885

[Composable API] `replicate`: add support for DDP args (#89243)

yhcharles committed 2 years ago

0f274ed3

[FSDP][optim_state_dict][1/N] Restructure _optim_state_dict to prepare the support of use_orig_param (#89898)

fegin committed 2 years ago

72fdfad4

Simplify by using yield from (#90160)

kit1980 committed 2 years ago

2b20a3d3

Fix issue 38095 TODOs in NCCL tests (#90033)

kit1980 committed 2 years ago

54858cce

[NNC] Use New PassManager for LLVM >= 15 (#89978)

davidberard98 committed 2 years ago

7571134f

Assume that co_firstlineno is always defined (#90180)

ezyang committed 2 years ago

5de5c5e4

workaround for indexing formulas with negative terms (#89933)

Natalia Gimelshein committed 2 years ago

1ea20cdb

fix c10::detail::integer_iterator for C++17 (#90174)

mikey dagitses committed 2 years ago

368a1cbd

Older