pytorch/xla

Pull Requests Commits

Add test for dynamo and sgd optimizer

JackCaoG committed 2 years ago

a7f036f4

[PJRT] Use the new local rank env var to calculate local_ordinal (#4684)

will-cromar committed 2 years ago

Verified c5d4103a

fix starting profiler server (#4689)

lsy323 committed 2 years ago

Verified 3b106635

[FSDP] Replace all .data usages with modern approaches (#4485)

alanwaketan committed 2 years ago

Verified 5e6a32d0

Update other release regex patterns (#4685)

steventk-g committed 2 years ago

Verified a0eb5c8f

Fix diff filepath (#4683)

steventk-g committed 2 years ago

Verified 58894f0e

std/var: Add support for floating point correction (#4674)

wonjoo-wj committed 2 years ago

Verified f0b1ba11

Use Ansible for building wheels and provisioning docker images. (#4531)

mateuszlewko committed 2 years ago

Verified 8b3fc227

Fix torch-xla MSELoss to handle mixed datatype, also reuse the IR datatype instead of converting to input's. (#4672)

aws-bowencc committed 2 years ago

Verified 8ac9159d

Add regex matching for release branches (#4676)

steventk-g committed 2 years ago

Verified cd769519

[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

Older