strisunshinewentingwang/pytorch

Pull Requests Commits

[dynamo] Improve error message for missing backend (#97255)

jansel committed 1 year ago

de2230ba

Fix typo in settings regex logging (#97245)

ezyang committed 1 year ago

ec3894ec

Refactor NT offsets metadata to be a Tensor (#96909)

jbschlosser committed 1 year ago

77e73b9b

Change 1D Tensor of 1 element to 0D Tensor (#96994)

crcrpar committed 1 year ago

22ea21da

Update parallel_apply.py for assertion error when len(modules) != len(inputs) (#94671)

qkaTlehdrnf committed 1 year ago

c47cf9bc

Fix required template (#97247)

qqaatw committed 1 year ago

a6bbeec2

Fix the compatible issue of the Dynamo and the PyDev.Debugger. (#96721)

chengjunlu committed 1 year ago

dbb31672

[CI] Fix perf_nightly output file naming error (#97263)

desertfire committed 1 year ago

b95896c5

[inductor] Add scaled_dot_product_attention to fallback kernels (#93339)

bertmaher committed 1 year ago

acd9df8a

Update mkl_verbose return value check due to API change in mkl (#96283)

jingxu10 committed 1 year ago

0a2b527a

Mark ROCm tests as flaky (#97259)

huydhn committed 1 year ago

244736a5

Make split reduction warning only emit once (#97112)

ezyang committed 1 year ago

5d3c347b

FIX make sure we import the correct object from multiprocessing (#81862)

tomMoral committed 1 year ago

701cdbb6

Fix uniform returning end point for BFloat16 and Half (#96962)

peterbell10 committed 1 year ago

4e054175

Run functorch tests in default shards; delete functorch-specific shards (#96464)

zou3519 committed 1 year ago

5acf4030

Re-enable TestJit.test_profiler (#94391)

mantaionut committed 1 year ago

b004819f

Allow new_full's fill_value argument type to be complex (#91345)

ccshan committed 1 year ago

2c588b3a

[PTD][Checkpoint] Add checkpointing support for DTensor submesh (#96802)

wz337 committed 1 year ago

38b687ed

[Inductor] index_put - unsqueeze indices[0] if self and indices[0] are not broadcastable (#97105)

yanboliang committed 1 year ago

a9b9fd90

Clean up Compilation Profiler (#97029)

wconstab committed 1 year ago

141a2ebc

Extend aot autograd dedup guards to params, stop using positions (#96774)

voznesenskym committed 1 year ago

f9ce5932

[Quant][FX] Fix issue of lowering weighted functional ops with kwargs (#95865)

Xia-Weiwen committed 1 year ago

e8be6d81

[PyTorch] Remove unnecessary unpickler.h #include in jit/serialization/import.h (#96687)

swolchok committed 1 year ago

7beac103

[ONNX] 'Transform' as base class for passes (#95935)

BowenBao committed 1 year ago

d2f57229

Fix for verify_dynamo on ROCm (#97013)

jataylo committed 1 year ago

45296f87

Error only if autocast actually enabled (#96097)

HennerM committed 1 year ago

ee6b19bd

[inductor] Move fx-fusion tests to a separate file (#97028)

bertmaher committed 1 year ago

cc0701e5

[inductor] Allow `tensors` kwarg in sink_cat_after_pointwise (#97019)

bertmaher committed 1 year ago

695d98b0

[RFC] Add an API to remove autograd hooks from DDP (#96490)

pritamdamania87 committed 1 year ago

e20e5f55

Don't run fallback if symbolic sizes in fake tensor (#97148)

ezyang committed 1 year ago

fa820800

Newer Older