strisunshinewentingwang/pytorch

Pull Requests Commits

[Dynamo] Supports calling parent class‘s non classmethod from child class (#90682)

yanboliang committed 1 year ago

e2674aaf

[ao] fix incorrect integer cast on histogram observer bounds (#90355)

HDCharles committed 1 year ago

e1165088

Better url in trymerge (#90583)

clee2000 committed 1 year ago

60e196c2

[ONNX] Add repro export from `GraphInfo` (#89947)

BowenBao committed 1 year ago

f2587537

[ONNX] Verification tool to find mismatch in model export (#89946)

BowenBao committed 1 year ago

525c33c0

fix with statement in test_fsdp_hybrid_shard.py (#90580)

jeffdaily committed 1 year ago

4ed175bf

[optim] skip .item calls in all optimizers when compiling with dynamo (#88173)

soumith committed 1 year ago

06326a77

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

Skylion007 committed 1 year ago

7541c9f8

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

peterbell10 committed 1 year ago

27932ff8

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

Newer Older