strisunshinewentingwang/pytorch

Pull Requests Commits

s/dynamo.optimize/torch.compile

wconstab committed 1 year ago

f4321e12

Fix broken docs for CompileProfiler

wconstab committed 1 year ago

14e8ec02

Run bazel jobs on 4xlarge (#92340)

huydhn committed 1 year ago

b26efd0d

[optim][rprop] group tensors in foreach to maximize perf (#92372)

janeyx99 committed 1 year ago

bb34461f

Reclassify some dynamic aot_eager failures as static failures (#92376)

ezyang committed 1 year ago

b92a7afe

Fix and update type hints for `make_functional.py` (#91579)

XuehaiPan committed 1 year ago

ae4ec7de

Remove setup-python on Windows CI and use Conda instead (#92183)

huydhn committed 1 year ago

bcd9f189

Update clang-tidy to 15.0.6 (#92195)

malfet committed 1 year ago

65056845

[BE] Delete unused args during docker build (#92396)

malfet committed 1 year ago

74bc894e

Revert "Improve hooks ordering behavior (#85849)"

pytorchmergebot committed 1 year ago

e525f433

Add missing gc untrack for cpp autograd Nodes (#92351)

albanD committed 1 year ago

7f0d321d

[BE][optim] abstract out docstrings, add differentiable docs (#92336)

janeyx99 committed 1 year ago

0070c546

Allow DDP to handle custom dataclass forward outputs (#92334)

mrshenli committed 1 year ago

00353404

Deprecate torch.nn.utils.stateless.functional_call (#92280)

zou3519 committed 1 year ago

5d01277f

Add deprecation messages for functorch.* function transforms (#92279)

zou3519 committed 1 year ago

a8a44a1a

stack_module_state should return unrelated parameters (#92278)

zou3519 committed 1 year ago

21d2bd78

[dynamo] exclude reset_rng_state when measure timing (#92237)

chunyuan-w committed 1 year ago

3aa6cec1

Make masked_fill reference traceable (#90972)

peterbell10 committed 1 year ago

f0b592da

[PT-D][5/N] Enable add_param_group for named optimizer (#91928)

fduwjj committed 1 year ago

368c7376

[Quant][Eager] Copy MHA's batch_first attribute in prepare() (#91680)

Xia-Weiwen committed 1 year ago

61a7618f

Replace exp(x) - 1 with expm1(x) (#92154)

peterbell10 committed 1 year ago

206f4e47

Replace log(1 + x) with log1p(x) (#92114)

peterbell10 committed 1 year ago

4058dedf

[Quant] onednn backend switch to ideep new api without affacting performance (#91056)

Xia-Weiwen committed 1 year ago

5a2ae880

[Inductor] added aten.exponential_ decomp (#91673)

min-jean-cho committed 1 year ago

fb50a4b4

Retire unsafe sparse tensor constructors in Python API (#91331)

pearu committed 1 year ago

4a4520e7

Clean up C++14 code (#92216)

cyyever committed 1 year ago

dfbdfb27

[dtensor][3/N] move OpSchema and types to a separate file (#90732)

wanchaol committed 1 year ago

c55f6973

[dtensor][2/N] add __repr__ to placements (#91785)

wanchaol committed 1 year ago

dc95ef25

[dtensor][1/N] add __hash__ to device_mesh and dtensor_spec (#90731)

wanchaol committed 1 year ago

a1186d6a

Clear references to user tensors after compilation is finished (#92353)

mlazos committed 1 year ago

bc9af74c

Older