strisunshinewentingwang/pytorch

Pull Requests Commits

[reland3][dynamo] Revert "Revert "[reland][dynamo] use optimizers correctly in benchmar… (#90956)

mlazos committed 3 years ago

7c524221

Revert "[inductor] add conv_transpose2d unary fusion for cpu in inference mode (#90265)"

pytorchmergebot committed 3 years ago

78efde92

[Quant][FX] Add backend config for onednn backend and fuse Linear-LeakyReLU (#88665)

Xia-Weiwen committed 3 years ago

7b0ec67e

[Checkpoint] Fix checkpoint test test_fsdp_optim_state.py (#91036)

wz337 committed 3 years ago

bfa223aa

Remove duplicate line (#91006)

ZainRizvi committed 3 years ago

1d948787

[reland][quant][pt2e] Add early prototype top level quantize_pt2e APIs (#91035)

jerryzh168 committed 3 years ago

f7b384cc

fix default partitioner: save sizes instead of tensor for backward when possible (#91012)

bdhirsh committed 3 years ago

4ab81ae8

Save and restore tracked_fakes (#90995)

ezyang committed 3 years ago

1609b954

[functorch] add composition-of-3-transform tests for autograd_function (#90962)

zou3519 committed 3 years ago

ed589dd8

Fix comment about get_fw_grad_mode() only being used in custom Function (#90790)

zou3519 committed 3 years ago

e1c799ff

Add VmapInterpreter.randomness (in pyfunctorch) provide it in info object (#90789)

zou3519 committed 3 years ago

ffa37c9f

set -Winconsistent-missing-override for builds (#89851)

mikey dagitses committed 3 years ago

8bd959e4

lint transformer.py (#91048)

davidberard98 committed 3 years ago

93cb5806

[dynamo] turn torch.backends.cudnn.is_acceptable into a constant (#90323)

davidberard98 committed 3 years ago

5d70d128

Revert "add conv_transpose2d pointwise(unary) fusion kernel (#90264)"

pytorchmergebot committed 3 years ago

7d3f2b79

Allow Process Group to support multiple backends (#88330) (#90997)

H-Huang committed 3 years ago

7a0f29b7

[dynamo] Refactor how autocast parameters are binded (#90953)

desertfire committed 3 years ago

93ac8c4a

Add macro C10_AS_INTARRAYREF_SLOW (#90675)

ezyang committed 3 years ago

Verified 4fa8d774

Revert "[cuDNN][cuDNN V8 API] (re-re-open) cuDNN V8 API on by default (#89022)"

pytorchmergebot committed 3 years ago

ba7aeac3

Fix non-existing parameters in docstrings in torch/ao (#90875)

kit1980 committed 3 years ago

4438b019

ModuleInfo-based tests for AOTAutograd (#90980)

jbschlosser committed 3 years ago

ee247586

LSTM SymInt-aware changes & meta registration (cuDNN) (#90944)

jbschlosser committed 3 years ago

32262096

Introduce causal mask (#90508)

Michael Gschwind committed 3 years ago

512ec181

Don't recompute var in bn decomp (#90984)

Natalia Gimelshein committed 3 years ago

e689c509

functorch: non-contig samples for test_grad (#90990)

kshitij12345 committed 3 years ago

e4de6ed6

[FSDP][3/N] Move `fsdp_modules(root_only=True)` -> `_get_fsdp_root_states()` (#90862)

Andrew Gu committed 3 years ago

5ea418bf

Revert "[Quant] onednn backend switch to ideep new api without affacting performance (#90354)"

pytorchmergebot committed 3 years ago

67ef88af

aot_autograd: add assert for functional-only graph (#88816)

bdhirsh committed 3 years ago

7a683eae

[GHA][BE] Update to newer checkout action (#90969)

malfet committed 3 years ago

c83ff1ea

[quantized] [executorch] typo (#89960)

JacobSzwejbka committed 3 years ago

bd94ee66

Newer Older