strisunshinewentingwang/pytorch

Pull Requests Commits

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

srossross committed 1 year ago

dcce5677

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

pytorchmergebot committed 1 year ago

e37c8c84

Don't enable debug_fake_crossref for TORCH_COMPILE_DEBUG (#90666)

ezyang committed 1 year ago

0b3316ad

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

ezyang committed 1 year ago

f7365eca

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

pytorchmergebot committed 1 year ago

67023454

Shape guard structure (#90679)

voznesenskym committed 1 year ago

5adc18dc

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

yanboliang committed 1 year ago

2e0ce248

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

ppwwyyxx committed 1 year ago

c8ed84ad

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

XiaobingSuper committed 1 year ago

4ca2fc48

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

bertmaher committed 1 year ago

c318de42

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

bertmaher committed 1 year ago

b95ea4f1

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

bertmaher committed 1 year ago

d3d85e1c

Preserve original GraphArgs for shape guard codegen (#90665)

ezyang committed 1 year ago

8fd31ac4

Improve dynamo debug logging (#90664)

ezyang committed 1 year ago

9447005a

Slightly improve error messages on sympy failure (#90655)

ezyang committed 1 year ago

450bd282

Skip some unittests (#90609)

ppwwyyxx committed 1 year ago

8127724c

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

voznesenskym committed 1 year ago

11442acc

Add a timeout to benchmark script (#90634)

ezyang committed 1 year ago

e1ed5ad5

Some memory saving in large unittests (#90148)

ppwwyyxx committed 1 year ago

5d8618df

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

Skylion007 committed 1 year ago

995d39c2

SymIntify resize_ and deduplicate memory format logic (#90442)

ezyang committed 1 year ago

e33f1eee

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

jgong5 committed 1 year ago

181d3747

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

pytorchmergebot committed 1 year ago

15a4c603

[FSDP] Fix _pre_forward type annotation (#90621)

mrshenli committed 1 year ago

7ec1cb85

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

mrshenli committed 1 year ago

80542add

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

awgu committed 1 year ago

31351c61

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

voznesenskym committed 1 year ago

933b6c4e

Adopt state_dict_pre_hook in FSDP (#90436)

rohan-varma committed 1 year ago

c7d2fb7f

[FSDP][Easy] Move to `_storage()` in test file (#90622)

awgu committed 1 year ago

746c773d

[FSDP] Uncomment test for `use_orig_params=True` (#90610)

awgu committed 1 year ago

68455986

Newer Older