pytorch/xla

Pull Requests Commits

fix CI and add some debugging messages

JackCaoG committed 2 years ago

69c8dde9

JackCaoG committed 2 years ago

24f1a322

Move where clear pending IR is called to avoid crash

JackCaoG committed 2 years ago

e2935967

Make cpu tensor on XLA dynamo backend a warning instead of error (#5549)

JackCaoG committed 2 years ago

Verified e51d28b6

Add nightly wheel for py3.10 and cuda 11.8 (#5547)

vanbasten23 committed 2 years ago

Verified f38e4a50

Fix pip install command typo in GPU doc. (#5545)

vanbasten23 committed 2 years ago

Verified 2c8ebd14

Update nightly to 2.2 (#5542)

JackCaoG committed 2 years ago

Verified 04118d5e

Allow to override PyTorch git revision and set it to v2.1.0-rc2 for 2.1 releases (#5543)

mateuszlewko committed 2 years ago

Verified 3f978dfb

Add option to unbundle libtpu (#5534)

will-cromar committed 2 years ago

Verified d757e85e

Fix FSDP for Models with Frozen Weights (#5484)

Liyang90 committed 2 years ago

Verified 3bdb4b10

Fix some typos (#5521)

apivovarov committed 2 years ago

Verified b8b0e716

Add doc for stablehlo (#5523)

qihqi committed 2 years ago

Verified e22c9ce1

[OpenXLA] Add openxla.md (#5507)

ManfeiBai committed 2 years ago

Verified 02c4aff4

Add PyTorch/XLA SPMD README.md (#5490)

yeounoh committed 2 years ago

Verified 4582675b

Enable megacore_dense by default (#5520)

alanwaketan committed 2 years ago

Verified 17a0b7c0

Modify kokoro to make it work (#5457)

qihqi committed 2 years ago

Verified 17513f74

Add dtype check to amp for tpu (#5505)

cowanmeg committed 2 years ago

Verified a06b1dff

[SPMD] Propagate replicated output (#5508)

alanwaketan committed 2 years ago

Verified 5c4ce2a9

mark_sharding over a replicated tensor is allowed. (#5513)

yeounoh committed 2 years ago

Verified 6e90569d

Get original_traced_args as example_inputs. (#5511)

qihqi committed 2 years ago

Verified 33f1cd29

Update dynamo.md (#5509)

shauheen committed 2 years ago

Verified 808015d8

Update dynamo doc (#5506)

JackCaoG committed 2 years ago

Verified 2f4ef46d

Fix tupled partition spec test on v3 (#5503)

jonb377 committed 2 years ago

Verified 81063fb8

remove torchvision in stablehlo.py (#5501)

lsy323 committed 2 years ago

Verified 1c42ba74

Add explict error message when tensor is on CPU for dynamo backend (#5499)

JackCaoG committed 2 years ago

Verified 9557fb72

Add a API to explictly init runtime (#5500)

JackCaoG committed 2 years ago

Verified e1ee592d

Support tuples in partition spec (#5488)

jonb377 committed 2 years ago

Verified 7a63c276

Refactor stablehlo API and put them in official location. (#5493)

qihqi committed 2 years ago

Verified 17abce76

Skip a couple tests on TPU due to precision issue (#5496)

JackCaoG committed 2 years ago

Verified d3d6ca30

Always use ExecuteReplicated with SPMD (#5494)

jonb377 committed 2 years ago

Verified 03991d44

Older