pytorch/xla

Pull Requests Commits

optimization barrier

alanwaketan committed 1 year ago

d6cba709

alanwaketan committed 2 years ago

4afa5d50

Introduce LIBTPU_INIT_ARGS

alanwaketan committed 2 years ago

c7f477b7

Adds more comment

alanwaketan committed 2 years ago

a0e0d3d8

Set --xla_latency_hiding_scheduler_rerun to 1

alanwaketan committed 2 years ago

90d87741

Revert "Don't set $TPU_LIBRARY_PATH during import (#5698)" (#5731)

alanwaketan committed 2 years ago

Verified 4baef3c6

Promote in convolution (#5727)

qihqi committed 2 years ago

Verified 47bf2eac

Fix the missing parameter error when running mp_imagenet with torchrun (#5729)

vanbasten23 committed 2 years ago

Verified 294610a1

Make the git version appears in pip (#5728)

alanwaketan committed 2 years ago

Verified ecc0e230

Ensure dist runtime is in sync before shutting down. (#5714)

vanbasten23 committed 2 years ago

Verified fa7828f5

Patch RNG for better memory utilization in dropout layers (#5710)

yeounoh committed 2 years ago

Verified 8f121c31

add op lowering for _prelu_kernel_backward (#5724)

zpcore committed 2 years ago

Verified d28b30d2

Make sure printing XLA tensor only execute the HLO once (#5721)

JackCaoG committed 2 years ago

Verified 1ad310b7

Update OpenXLA pin to 20231022 (#5720)

alanwaketan committed 2 years ago

Verified 43cd6b50

When exporting StableHLO to SavedModel, also include the original var (#5711)

haozha111 committed 2 years ago

Verified e7af313d

Remove GetTensorsFused since we don;t have opbyop anymore (#5718)

JackCaoG committed 2 years ago

Verified 5dfcec90

fix typo in TROUBLESHOOTING (#5716)

JackCaoG committed 2 years ago

Verified 4fa727b4

Run decomp before processing (#5713)

qihqi committed 2 years ago

Verified b1123dbb

Add support for `_unsafe_index`. (#5707)

ysiraichi committed 2 years ago

Verified bccbb5aa

Add multi-host GPU support (#5657)

vanbasten23 committed 2 years ago

Verified 6ea99474

Update Troubleshooting with some sanity check example (#5705)

JackCaoG committed 2 years ago

Verified 8f45caef

Filter tensor arguments from traced model. (#5689)

ysiraichi committed 2 years ago

Verified 994f9fb6

pass bundle object to make_tf_function (#5708)

haozha111 committed 2 years ago

Verified 126c9b15

Don't set $TPU_LIBRARY_PATH during import (#5698)

will-cromar committed 2 years ago

Verified 146f2a01

Add --net=host and --shm-size=16g flag to the docker run command in GPU doc. (#5702)

vanbasten23 committed 2 years ago

Verified c9aa49c3

Support async checkpointing through CheckpointManager (#5697)

jonb377 committed 2 years ago

Verified 06ba6e92

Reduce unnecessary tensor allocation in Adam and AdamW (#5700)

baoleai committed 2 years ago

Verified d9523716

Support synchronous saving and loading in CheckpointManager (#5693)

jonb377 committed 2 years ago

Verified 2a45d0d6

Update CI image with dev container image (#5290)

lsy323 committed 2 years ago

Verified e793e864

Open XLA pin update (#5675)

qihqi committed 2 years ago

Verified 418c7518

Older