pytorch/xla

Pull Requests Commits

[Dist] Add ZeRO-1 Optimizer (#4648)

hgt312 committed 2 years ago

Verified 1d313bb0

Use GCC and remove sandboxing mechanisms. (#4658)

stgpetrovic committed 2 years ago

Verified 023d763b

Run single device for GPU coverage (#4667)

jonb377 committed 2 years ago

Verified 35c07d52

[PJRT] Change the name of `LOCAL_RANK` to fix conflicts (#4646)

will-cromar committed 2 years ago

Verified 54101e76

Add spmd unit tests and linear model training to CI (#4645)

steventk-g committed 2 years ago

Verified a2b673f0

Iron out the cuda/tpuvm flag (#4657)

stgpetrovic committed 2 years ago

Verified 2d9b37b9

Revert to local nonsandboxed builds with GCC (#4651)

stgpetrovic committed 2 years ago

Verified e6c593b9

Full codegen for repeat (#4642)

cowanmeg committed 2 years ago

Verified 4ea000b9

Add experimental test coverage feature (#4316)

yeounoh committed 2 years ago

Verified a8ac0d91

Make WaitDeviceOps block until device execution finishes (#4626)

JackCaoG committed 2 years ago

Verified 6d1158cf

Disable C++ ops tests failing on XRT TPU (#4643)

will-cromar committed 2 years ago

Verified a7b0d52a

Disable docker sandbox (#4649)

stgpetrovic committed 2 years ago

Verified 3f091bdd

Update the test to use the more standard torch.compile api (#4634)

JackCaoG committed 2 years ago

Verified b862368c

[PJRT] Defer default device selection until we need to know the runtime (#4630)

will-cromar committed 2 years ago

Verified 32da64f6

stgpetrovic committed 2 years ago

Verified 1bbe4da7

Revert "Bazel (#4632)" (#4635)

stgpetrovic committed 2 years ago

Verified b98179fc

stgpetrovic committed 2 years ago

Verified 426c5cde

Revert "Bazel (#4528)" (#4631)

JackCaoG committed 2 years ago

Verified 286ef02d

stgpetrovic committed 2 years ago

Verified 3a90f5af

Disable failing `test_operations.py` tests on TPU (#4622)

will-cromar committed 2 years ago

Verified 83f425a1

Skip execution for extract_compiled_graph (#4612)

JackCaoG committed 2 years ago

Verified b7707fa1

Fix patch command in TPU CI (#4623)

will-cromar committed 2 years ago

Verified 574cfdab

Update TF pin to 2/13 (#4615)

will-cromar committed 2 years ago

Verified dbe596c7

Fix HLO dumping (#4619)

JackCaoG committed 2 years ago

Verified f70afa45

Remove reference to `torch._six` compat layer (#4613)

XuehaiPan committed 2 years ago

Verified d29eb67c

[FSDP] Use state_dict hook to save shard metadata for consolidating checkpoint (#4572)

YangFei1990 committed 2 years ago

Verified d13090c4

Support multihost SPMD execution (#4573)

jonb377 committed 2 years ago

Verified ced64569

[PJRT] Use TFRT TPU client by default (#4600)

will-cromar committed 2 years ago

Verified 16b4f03b

Add int as supported index tensor type (#4602)

ymwangg committed 2 years ago

Verified 249b8e69

Support multiple GPUs on a single host with PjRt (#4297)

jonb377 committed 2 years ago

Verified 1e011c60

Newer Older