pytorch/xla

Pull Requests Commits

Increase the default XLA_IR_SHAPE_CACHE_SIZE to accomunate the larger model size

JackCaoG committed 2 years ago

0cf761e8

Disable cxx abi in the release wheel. (#4900)

vanbasten23 committed 2 years ago

Verified 489a8ff9

Use dynamic shape support in TFRT TPU and CPU (#4921)

will-cromar committed 2 years ago

Verified 21970512

[Distributed] Make xm.all_gather a single graph in Dynamo (#4922)

alanwaketan committed 2 years ago

Verified 486b32e5

[PJRT] Add `tpu.num_available_devices` to get number of XLA TPU devices (#4927)

will-cromar committed 2 years ago

Verified ef95cb41

use test data for test loader (#4924)

vanbasten23 committed 2 years ago

Verified 045aa264

Remove rendezvous from xm.save (#4926)

will-cromar committed 2 years ago

Verified 4f7f3dcd

[Distributed] Test if xm.all_reduce produces a single dynamo graph (#4920)

alanwaketan committed 2 years ago

Verified 08574298

[Distributed] Adopts traceable all_reduce (#4915)

alanwaketan committed 2 years ago

Verified 6c7ca993

Update op lowering guide (#4925)

cowanmeg committed 2 years ago

Verified 3c00d573

[Distributed] Hide token from Python for _xla_all_reduce path (#4913)

alanwaketan committed 2 years ago

Verified 4994597f

[Distributed] Move the cached all reduce token to C++ (#4912)

alanwaketan committed 2 years ago

Verified 44c2fa04

[PJRT] Make `rendezvous` optional in `xm.save` (#4907)

will-cromar committed 2 years ago

Verified 3f8abbdc

Apply sharding to all compatible input tensors (#4919)

jonb377 committed 2 years ago

Verified a537020f

Skip test/test_zero1.py for TPU (#4917)

alanwaketan committed 2 years ago

Verified b4b4d0dc

Faster multinomial implementation (#4904)

cowanmeg committed 2 years ago

Verified b1862ccc

Revert "Apply sharding to all compatible input tensors (#4872)" (#4911)

alanwaketan committed 2 years ago

Verified fa059c47

[Distributed] Remove sea-of-devices mode (#4905)

alanwaketan committed 2 years ago

Verified 397ea4d5

Add a test to verify nonzero correctness. (#4903)

vanbasten23 committed 2 years ago

Verified 8af5207d

Pin setuptools version to fix vision build (#4906)

JackCaoG committed 2 years ago

Verified 1e54c235

Update tf pin to 03/2023 (#4840)

wonjoo-wj committed 2 years ago

Verified c1138e88

Lower linalg_vector_norm (#4781)

cowanmeg committed 2 years ago

Verified 11c4d2bd

Enable PJRT C API Client for Intel XPU (#4891)

jzhoulon committed 2 years ago

Verified baea8c55

Update dynamo readme (#4759)

JackCaoG committed 2 years ago

Verified 936b3b84

Apply sharding to all compatible input tensors (#4872)

jonb377 committed 2 years ago

Verified 919f0aee

Update 2.0 to 2.1 in experimental Docker (#4898)

will-cromar committed 2 years ago

Verified 855ba3c3

Update PJRT readme to point to r2.0 during clone (#4894)

JackCaoG committed 2 years ago

Verified 080a70d2

Fix formatting in `xla_client` (#4892)

will-cromar committed 2 years ago

Verified dc973296

Add html to C++ coverage report (#4886)

steventk-g committed 2 years ago

Verified c98ce753

Lower view_as_complex_copy and view_as_real_copy (#4887)

wonjoo-wj committed 2 years ago

Verified a47a7818

Older