strisunshinewentingwang/pytorch

Pull Requests Commits

[Fix]: remove unnecessary copies in aten, c10, and torch bindings (#90629)

Skylion007 committed 2 years ago

7541c9f8

[Inductor] Add note that stride_vars result may be inaccurate (#90184)

peterbell10 committed 2 years ago

27932ff8

Adding test when registering a batching rule for a CompositeImplicitAutograd operation (#89465)

srossross committed 2 years ago

dcce5677

Revert "[inductor] Update TIMM skip list (#90188)"

pytorchmergebot committed 2 years ago

e37c8c84

Don't enable debug_fake_crossref for TORCH_COMPILE_DEBUG (#90666)

ezyang committed 2 years ago

0b3316ad

Add unbacked symints support; item works now (#90624)

ezyang committed 2 years ago

f7365eca

[xla hash update] update the pinned xla hash (#90161)

pytorchmergebot committed 2 years ago

67023454

Shape guard structure (#90679)

voznesenskym committed 2 years ago

5adc18dc

[Dynamo] Support access nn.Module keys (#90502)

yanboliang committed 2 years ago

2e0ce248

Fix a static initialization order fiasco in c10d (#90149)

ppwwyyxx committed 2 years ago

c8ed84ad

inductor(CPU): add Conv+binary+unary fusion filter (#90259)

XiaobingSuper committed 2 years ago

4ca2fc48

[dynamo] Get GPU names without calling nvidia-smi (#90474)

bertmaher committed 2 years ago

c318de42

[pt2] Reset dynamo log level when exiting inductor debug context (#90473)

bertmaher committed 2 years ago

b95ea4f1

Emit torch.cuda.synchronize() after every kernel call in inductor (#90472)

bertmaher committed 2 years ago

d3d85e1c

Preserve original GraphArgs for shape guard codegen (#90665)

ezyang committed 2 years ago

8fd31ac4

Improve dynamo debug logging (#90664)

ezyang committed 2 years ago

9447005a

Slightly improve error messages on sympy failure (#90655)

ezyang committed 2 years ago

450bd282

Skip some unittests (#90609)

ppwwyyxx committed 2 years ago

8127724c

Make torch._guards, shuffle structures around for migration (#90636)

voznesenskym committed 2 years ago

11442acc

Add a timeout to benchmark script (#90634)

ezyang committed 2 years ago

e1ed5ad5

Some memory saving in large unittests (#90148)

ppwwyyxx committed 2 years ago

5d8618df

[Fix]: Add some missing moves in 90442 (#90661)

Skylion007 committed 2 years ago

995d39c2

SymIntify resize_ and deduplicate memory format logic (#90442)

ezyang committed 2 years ago

e33f1eee

Simple fix: add missing positional arg in init_optimizer() call (#90641)

jgong5 committed 2 years ago

181d3747

Revert "Make torch._guards, shuffle structures around for migration (#90636)"

pytorchmergebot committed 2 years ago

15a4c603

[FSDP] Fix _pre_forward type annotation (#90621)

mrshenli committed 2 years ago

7ec1cb85

[FSDP] Allow MixedPrecision to skip inputs (#90620)

mrshenli committed 2 years ago

80542add

[FSDP] Tighten post-bwd cast to `reduce_dtype` (#90615)

awgu committed 2 years ago

31351c61

Make torch._guards, shuffle structures around for migration (#90636)

voznesenskym committed 2 years ago

933b6c4e

Adopt state_dict_pre_hook in FSDP (#90436)

rohan-varma committed 2 years ago

c7d2fb7f

Newer Older