strisunshinewentingwang/pytorch

Pull Requests Commits

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

Re-enable Windows debug libtorch

malfet committed 2 years ago

bebfdca0

[quant][fix] Fix bug for ave pooling in FX quant (#73054)

terrychenism committed 2 years ago

5167e9d5

Use higher timeout for test_tensorpipe_set_default_timeout (#73771)

pritamdamania committed 2 years ago

aca4d02d

Lock thread/block computation (#73800)

Elias Ellison committed 2 years ago

52ccbf44

[ONNX] Rewrite linspace symbolic

BowenBao committed 2 years ago

4a74285e

Revert D34554432: Back out "Revert D34524207: [pytorch][PR] remove _s_where"

ngimel committed 2 years ago

55525632

[PyTorchEdge] Add getFileFormat() so we can differentiate Zip/Pickle from Flatbuffer (#73707)

dbort committed 2 years ago

7b51629c

[FSDP][BE] s/ConfigAutoWrap/_ConfigAutoWrap (#73759)

rohan-varma committed 2 years ago

0cbb458b

Fix typo in error message (THCGenerateByteType.h) (#73670)

yulonglin committed 2 years ago

03484828

Back out "Revert D34524207: [pytorch][PR] remove _s_where" (#73579)

ngimel committed 2 years ago

9c03c616

[PyTorch] Make NestedTensor::dim() work (#73679)

swolchok committed 2 years ago

88b19da4

[FX] Make `immutable_list` and `immutable_dict` work with pytrees (#73766)

James Reed committed 2 years ago

a8d9fbb0

[FX lowering] Modify replace_all_uses_with to allowing filtering of nodes to update; use it to (#73763)

jaybean-dev committed 2 years ago

5332d870

Fix `functional.max_poolNd` warning spam in the CI

krshrimali committed 2 years ago

452c26bb

[quant][fx] Refactor _convert_fx_do_not_use (#73777)

jerryzh168 committed 2 years ago

8b8fac91

[Model Averaging] Support hierarchical model averaging (#73285)

wayi1 committed 2 years ago

0bb3b065

[torch.distributed][DDP] Disable DDP bucketing for the first iteration (#72843)

xush6528 committed 2 years ago

bcd0843b

dbr quant: enable reference module support for torch.qint32 (#73493)

vkuzo committed 2 years ago

727debb1

dbr quant: insert activation obs explicitly, instead of relying on hooks (#73492)

vkuzo committed 2 years ago

5787a36e

Automated submodule update: FBGEMM (#73277)

facebook-github-bot committed 2 years ago

350ecc9d

[AutoAccept][Codemod][FBSourceClangFormatLinter] Daily `arc lint --take CLANGFORMAT`

CodemodService FBSourceClangFormatLinterBot committed 2 years ago

6067054c

Revert D34599476: [Quant][test] Added test to check if fp16 packing->unpacking yields the same result as to(torch.float16).to(torch.float32)

albanD committed 2 years ago

8a47d9cf

[NS] Mark output logger impure to avoid being removed in acc tracer (#73745)

Haixin Liu committed 2 years ago

3042f0ce

Update fbcode symlinks for mkl-dnn ideep 2.5.2

Wei Wei committed 2 years ago

bd3db019

[quant][fx] Add lowering support for qat and fused convs (#73527)

jerryzh168 committed 2 years ago

f5c7e540

[Quant][fx] Added explicit entries for for functional and module conv&linear support into get_default_qconfig_dict&get_default_qat_qconfig_dict (#73528)

dzdang committed 2 years ago

a39e8e8f

Eliminate unused parameters in PyTorch (#73749)

r-barnes committed 2 years ago

72e4aab7

Newer Older