pytorch/xla

Pull Requests Commits

opt-barrier weights and buffers

alanwaketan committed 1 year ago

3df4855b

minor rebase fix

alanwaketan committed 1 year ago

8df72e91

Fix clip grad norm

alanwaketan committed 1 year ago

89224660

alanwaketan committed 1 year ago

88f48833

alanwaketan committed 1 year ago

3be4dc9d

simplify it a bit

alanwaketan committed 1 year ago

1595ecc3

Prove the backward is right

alanwaketan committed 1 year ago

525cd45b

alanwaketan committed 1 year ago

29411e36

alanwaketan committed 1 year ago

6f41ac1f

alanwaketan committed 1 year ago

3cdcdef9

alanwaketan committed 1 year ago

963eea70

Use torch.autograd.backward

alanwaketan committed 1 year ago

6dc20ba8

alanwaketan committed 1 year ago

d8a90285

[torch_xla2] Use `torch.Generator` for random number generation in `Environment` (#7018)

will-cromar committed 1 year ago

Verified 94fa9993

Update index for torch_xla apis (#7171)

JackCaoG committed 1 year ago

Verified 8102ad42

Run some of the examples in TPU CI (#7172)

JackCaoG committed 1 year ago

Verified 0d754931

isolate the PJRT dependency between torch_xla and torch_xla2 (#7144)

zpcore committed 1 year ago

Verified 0f9bff34

`mul`: convert inputs to result type. (#7130)

ysiraichi committed 1 year ago

Verified 7938bb5d

Unify how we use torch commit in the CI. (#7164)

vanbasten23 committed 1 year ago

Verified 89832c3d

Reuse DLDeviceType. (#7163)

vanbasten23 committed 1 year ago

Verified 09cfdbf4

Add a CI workflow for tests that requires pytorch CUDA. (#7140)

vanbasten23 committed 1 year ago

Verified 6fadbf5d

Update configuration.yaml (#7158)

JackCaoG committed 1 year ago

Verified 8471826a

[Doc] Update spmd.md for doc (#7019)

ManfeiBai committed 1 year ago

Verified cb482bca

Add optimizer priming for dist chkpt (#6572)

jonb377 committed 1 year ago

Verified 8fd051f2

add PT_XLA_DEBUG_LEVEL (#7149)

JackCaoG committed 1 year ago

Verified 8c2234ee

Make from_dlpack handle cuda synchronization implicitly for input tensors that have __dlpack__ and __dlpack_device__ attributes. (#7125)

vanbasten23 committed 1 year ago

Verified daada224

[Pallas] Introduce GMM(torch.autograd.Function) (#7152)

alanwaketan committed 1 year ago

Verified aeed61a9

Deprecate XLA_USE_BF16 (#7150)

JackCaoG committed 1 year ago

Verified af51f063

[Pallas] Introduce gmm_backward (#7151)

alanwaketan committed 1 year ago

Verified c96c95a4

[Pallas] Make repeat_with_fixed_output_size not OOM on VMEM (#7145)

alanwaketan committed 1 year ago

Verified ce1205e1

Older