strisunshinewentingwang/pytorch

Pull Requests Commits

[package] use WeakValueDictionary for global imported module registry (#51666)

suo committed 3 years ago

003a240e

Revert D26246231: [FX] Edits after comprehensive pass over docs

albanD committed 3 years ago

6c80fd00

Support at::cpu on non-structured kernels (#51590)

ezyang committed 3 years ago

4d85e301

Split anonymous and namespaced definitions in RegisterDispatchKey (#51585)

ezyang committed 3 years ago

668e0f35

Factor out structured generation into its own subclass. (#51583)

ezyang committed 3 years ago

a626b784

Split out RegisterDispatchKey to its own file (#51508)

ezyang committed 3 years ago

93c4f9f9

Use Literal to model targets. (#51500)

ezyang committed 3 years ago

6045663f

[FX] Edits after comprehensive pass over docs (#51705)

James Reed committed 3 years ago

c22bc482

Make te a hidden package for now (#51690)

albanD committed 3 years ago

9920ae66

Support Union[NoneType, T] as input type (#51605)

nikithamalgi committed 3 years ago

ecf81665

[profiler] Support top-level memory events (#51421)

Ilia Cherniavskii committed 3 years ago

f1f9b049

Fix attribution of some CUDA events to CPU events (#51632)

Ilia Cherniavskii committed 3 years ago

a9584f29

[profiler] Default activities value (#51561)

Ilia Cherniavskii committed 3 years ago

d6452a1a

add dumping callstack to kineto (#51565)

gaoteng-git committed 3 years ago

7abba67d

[Usability] Tolerate `torch.jit.script` call to Enum classes (#51624)

gmagogsfm committed 3 years ago

8c3e0ddb

Graceful invalidation of Python Node/Value/Block when C++ object is deleted (#50326)

t-vi committed 3 years ago

86861095

Expand benchmark utils docs (#51664)

Taylor Robie committed 3 years ago

c8af3384

unbreak bc test (#51702)

ngimel committed 3 years ago

1518aee6

Fix memory leak in qnnpack ops (#51612)

Akshit Khurana committed 3 years ago

6a945bfb

Generate header with version #defines for LibTorch (#50073)

jbschlosser committed 3 years ago

e60f18c2

[PyTorch Mobile] Support torchbind custom classes in lite interpreter (#51432)

iseeyuan committed 3 years ago

23c50a4a

[quant] Add reflection padding to conv (#49011)

z-a-f committed 3 years ago

1ffd26f8

Use deterministic impl of `index_put` and `index` backward CPU when `torch.are_deterministic_algorithms_enabled() == True` (#51388)

kurtamohler committed 3 years ago

c41678fd

[FX] Added how to write transformations section (#51278)

Chillee committed 3 years ago

f1a63b7c

Fixes cat backward formula to return correct gradient values for R -> C case (#51681)

anjali411 committed 3 years ago

bd3ae117

[quant] Support 2 dim input in quantized batchnorm 1d (#51597)

jerryzh168 committed 3 years ago

d8742eee

Fix caffee2 for LLVM trunk

htyu committed 3 years ago

5d123ecf

mem-efficient learnable fake quantization (#49315)

haichuan-fb committed 3 years ago

0c60922f

[FX] Move examples to pytorch/examples (#51686)

James Reed committed 3 years ago

7918f37e

Extend subgraph_rewriter logic (#51532)

Ansley Ussery committed 3 years ago

f2c4deab

Newer Older