strisunshinewentingwang/pytorch

Pull Requests Commits

reland "fix as_strided_scatter_backward (#87646)" (#88342)

bdhirsh committed 2 years ago

a16ced03

[Static Runtime] Fix tensor_split sections overload (#88113)

Mike Iovine committed 2 years ago

dd43903f

[xla hash update] update the pinned xla hash (#88070)

pytorchmergebot committed 2 years ago

7076a648

Support sign for HF models like ElectraForQuestionAnswering (#88160)

EikanWang committed 2 years ago

ad27d762

Support reduction vectorization (#87356)

EikanWang committed 2 years ago

a9d37ce8

Explicit vectorization support for TorchInductor (#87068)

EikanWang committed 2 years ago

6541e51f

use faster cache flush in triton benchmarking (#88557)

Natalia Gimelshein committed 2 years ago

a95419b4

[Dynamo] fix torchdynamo's TVM meta schedule backend (#88249)

YJ Shi committed 2 years ago

eda247ee

[inductor] Add lowering for as_strided_scatter (#88379)

peterbell10 committed 2 years ago

791d9ee2

Revert "Reenable optimizer overlap tests (#88439)"

pytorchmergebot committed 2 years ago

81042d3a

Add error inputs to `gaussian_nll_loss` `OpInfo` (#88486)

nkaretnikov committed 2 years ago

bbaa0637

Upstream apply_optim_in_backward from TorchRec (#87397) (#88539)

rohan-varma committed 2 years ago

404f254e

Reenable optimizer overlap tests (#88439)

rohan-varma committed 2 years ago

da452bca

Handle case when candidate is empty (#88359)

ezyang committed 2 years ago

d1ee0730

[Reland] Fix primTorch compute_elementwise_output_strides (#88525)

SherlockNoMad committed 2 years ago

46730aec

Functionalize and compute joint simultaneously. (#88063)

ezyang committed 2 years ago

0e3031f7

fx.replace_pattern accepts pattern/replacement as GraphModule (#88479)

SherlockNoMad committed 2 years ago

957a9b63

Add docstring to DDPOptimizer (#88521)

wconstab committed 2 years ago

4bb5c2c2

Add single-process DDP accuracy support to dynamo benchmark suite (#88511)

wconstab committed 2 years ago

1f32c3c0

DDPOptimizer replace debug=True/False with using torchdynamo logger (#88480)

wconstab committed 2 years ago

3fd0729b

nvprims native batch norm patch (#88455)

jjsjann123 committed 2 years ago

52375a0f

[Dynamo] Improve BuiltinVariable log when incorrect arg count happens (#88409)

yanboliang committed 2 years ago

b1116a51

Fix minifier accuracy msg (#88515)

mlazos committed 2 years ago

5220d07d

Populate self.export in InstructionTranslatorBase (#88508)

mergennachin committed 2 years ago

dde9affe

[QNNPACK] Add unaligned attributes where asan fails (#88276)

digantdesai committed 2 years ago

afdc2283

[Quant] Respect non_leaf_module_list for activation modules (#88498)

andrewor14 committed 2 years ago

7560a7b2

[BE] Update native_functions.yaml README; we do not support Tensor! (#88513)

janeyx99 committed 2 years ago

5af3feef

Support DDP ignored parameters in DDPOptimizer (#88460)

wconstab committed 2 years ago

678d0380

enable backward for log1p (sparse layouts) (#88155)

amjames committed 2 years ago

ff6770a9

Support sparse inputs to deg2rad (#88156)

amjames committed 2 years ago

6938dd0b

Newer Older