strisunshinewentingwang/pytorch

Pull Requests Commits

MHA forward pass bug fix

george-qi committed 2 years ago

f9db8b72

ezyang committed 2 years ago

e3403ff4

[FSDP] Use post load_state_dict hooks (#76912)

rohan-varma committed 2 years ago

4a57321a

masked cumsum/cumprod

george-qi committed 2 years ago

ea272443

Make ShufflerDataPipe deterministic for SP & MP DataLoader (#77741)

ejguan committed 2 years ago

365ce350

[shard] fix failed tests in sharded tensor

wanchaol committed 2 years ago

4124307f

[RELAND] Adds torch.cuda.is_current_stream_capturing (#77789)

mcarilli committed 2 years ago

929f1d53

Update scale-config.yml (#77803)

suo committed 2 years ago

0f328f35

Add testing workaround for EFA and TensorPipe (#77363)

Rodrigo Kumpera committed 2 years ago

dac3fba2

[GHF] Add URL for pending/failed mandatory checks (#77763)

malfet committed 2 years ago

d40a2400

Revert "Add sharding tests to multigpu-test.sh (#77708)"

pytorchmergebot committed 2 years ago

5e0f559d

[checkpoint] Synchronize error handling across all ranks (#77091)

Rodrigo Kumpera committed 2 years ago

c9570e4b

Support copy_ for Sparse Compressed tensors.

pearu committed 2 years ago

8b5f11c6

improve mps note to describe the different functions available (#77767)

albanD committed 2 years ago

dcd2ba35

Add USE_MPS option to cmake summary

drisspg committed 2 years ago

1f7d243e

Throw a nice error when SubTensor.__torch_dispatch__() returns the wrong type for detach()

jbschlosser committed 2 years ago

0794d59d

Support no-batch-dim for CrossEntropyLoss with prob target

jbschlosser committed 2 years ago

8881d7ac

rocblas alt impl during backward pass only (#71881)

jeffdaily committed 2 years ago

de86146c

Revert "Adds torch.cuda.is_current_stream_capturing (#77673)"

pytorchmergebot committed 2 years ago

0d8a0f18

Upstream master bump 0513 (#77471)

jjsjann123 committed 2 years ago

a2802ad0

Revert "Revert "Implement sym_sizes to create proper IR for sym ints representing tensor sizes (#76836)""

ezyang committed 2 years ago

4941e72e

ezyang committed 2 years ago

befa4e37

Fix 'Code below assumes there is at least one tensor arg' assumption (#76917)

antoniojkim committed 2 years ago

55be35ae

Add sharding tests to multigpu-test.sh (#77708)

pritamdamania87 committed 2 years ago

a7cf95a6

Adding a way to register both upper and lower bound functions

Gamrix committed 2 years ago

2a990181

Adding support for nonzero in LazyTensor shape functions

Gamrix committed 2 years ago

73480bcb

[fix] composite compliance : nuclear_norm (#77734)

kshitij12345 committed 2 years ago

23b2e882

[nvFuser] Improving bitwise ops support (#77158)

zasdfgbnm committed 2 years ago

4eec865f

[chalf] div, eq, masked_fill, index_put (#77479)

kshitij12345 committed 2 years ago

85710070

Adds torch.cuda.is_current_stream_capturing (#77673)

mcarilli committed 2 years ago

d03d43df

Older