strisunshinewentingwang/pytorch

Pull Requests Commits

clee2000 committed 2 years ago

ccf1fb00

clee2000 committed 2 years ago

08170663

[export] Port over const prop pass (#100102)

angelayi committed 2 years ago

7bece142

[PTD][Checkpoint] Upstream fsspec storage read/write to PT (#98387)

wz337 committed 2 years ago

fad2f6ed

[SPMD] Add embedding dense backward prop rule for postional embedding (#100038)

fegin committed 2 years ago

b94a0ba5

Remove CUDA 11.6 note from complex docs (#100118)

ptrblck committed 2 years ago

8fe91d16

Add private _export API (#99992)

tugsbayasgalan committed 2 years ago

02f059c2

[dynamo][numpy] Handle return value being numpy ndarray (#99560)

larryliu0820 committed 2 years ago

f5853342

[dynamo][numpy] Add NumpyTensorVariable to translate ndarray attribute calls to tensor attributes (#95849)

larryliu0820 committed 2 years ago

687afeb6

[Dynamo] Add unit test for explicitly calling __call__ (#100146)

yanboliang committed 2 years ago

d855b6ae

Fix cat forward-AD tests (#99596)

qqaatw committed 2 years ago

cb569dbc

[inductor] Fix argmin/max with duplicate values (#99920)

peterbell10 committed 2 years ago

659dcc5e

[inductor] Fix nan-handling of max and min reductions (#99881)

peterbell10 committed 2 years ago

f9c3fcd1

[inductor] Create triton_helpers module for helper functions (#99880)

peterbell10 committed 2 years ago

ed2eb13d

[c10d] Scalable PG initiation. (#99931)

Rodrigo Kumpera committed 2 years ago

ad21890f

sparse.sum backward: short circuit on zero/empty grad (#98838)

nikitaved committed 2 years ago

2eab5abb

Add support for serializing real tensor data in after aot minifier (#99834)

ezyang committed 2 years ago

67e0913d

relu/threshold backward for sparse: enable 0-nnz grads (#98935)

nikitaved committed 2 years ago

5cfaea15

Change caffe2 branch links to main (#100129)

kit1980 committed 2 years ago

c2402a92

Include all mkl/mkldnn related test files to CPU ATen backend (#99592)

Jiong Gong committed 2 years ago

77a37a54

Basic dynamo support for traceable collectives (#94440)

wconstab committed 2 years ago

100a25d0

[CUDA] Switch to `at::empty` in `max_pool3d_with_indices_backward_cuda` (#100138)

eqy committed 2 years ago

925a3788

Making fsdp device-agnostic for custom-backend which implement cuda-semantics (#99024)

medivh-xp committed 2 years ago

859e82a7

[inductor] fix _print_Pow given reciprocal of dynamic dim with float exponent (#100090)

Jiong Gong committed 2 years ago

4456e932

inductor: enhance conv+binary fusion path test for cpu path (#100058)

XiaobingSuper committed 2 years ago

569eff85

fix missing-prototypes warnings in torch_cpu (Part 1) (#100053)

cyyever committed 2 years ago

e2480164

[dynamo] Add ddp_graphs artifact (#100021)

bertmaher committed 2 years ago

e0bf51d3

[inductor] logger message fix in split_cat (#100088)

Jiong Gong committed 2 years ago

1504bdf9

Rename DispatchKey.PrivateUse1 to custom device in torchgen. (#99406)

huihoaan committed 2 years ago

c0ecd989

inductor: simplify the test_mkldnn_pattern_matcher.py code (#100057)

XiaobingSuper committed 2 years ago

3588688a

Older