pytorch/xla

Pull Requests Commits

Add tf independent serialization (#5308)

qihqi committed 2 years ago

Verified 0b5b5197

quick refactor on _get_group_assignment (#5318)

JackCaoG committed 2 years ago

Verified a8f723b6

Fix mismatched GPU docker image in the doc. (#5319)

vanbasten23 committed 2 years ago

Verified 42ae0bf9

Support unordered sharding spec for partial replication (#5316)

JackCaoG committed 2 years ago

Verified 78c6840c

Update gcc version in CI (#5297)

lsy323 committed 2 years ago

Verified e7e18996

Add docs for AMP (#5313)

cowanmeg committed 2 years ago

Verified 1cb5a524

Disable test failing on TPU CI (#5314)

JackCaoG committed 2 years ago

Verified cf491f9e

Fix race condition when use > 1 threads to transfer data in parallel loader (#5267)

aws-tianquaw committed 2 years ago

Verified 1dc5af55

Supoort unordered sharding spec correctly (#5305)

JackCaoG committed 2 years ago

Verified 67705065

Add `_sharded_cpu_state_dict` for distributed checkpointing (#5288)

shahyash10 committed 2 years ago

Verified 46a0117b

[SPMD] Allow dumping post optimizations hlo (#5302)

alanwaketan committed 2 years ago

Verified 15ee2191

[SPMD] Support mark_sharding on IRs (#5301)

alanwaketan committed 2 years ago

Verified 6fe5cb98

[SPMD] Avoid recompilations in xs.mark_sharding() (#5300)

alanwaketan committed 2 years ago

Verified caf51687

[SPMD] Add FSDP sharding for test_train_spmd_linear_model.py (#5299)

alanwaketan committed 2 years ago

Verified d01880c5

[SPMD] Use xs.Mesh in test_2d_tensor_3d_mesh (#5295)

khatwanimohit committed 2 years ago

Verified df680da7

Make README more actionable (#5262)

will-cromar committed 2 years ago

Verified c27a493e

Revert pr https://github.com/pytorch/xla/pull/2682 (#5215)

vanbasten23 committed 2 years ago

Verified 8cf85def

[SPMD][Virtual Device]All tensors should be in SPMD:0 C++ device (#5284)

JackCaoG committed 2 years ago

Verified e42fffa0

Fix typo in _test.yml (#5172)

malfet committed 2 years ago

Verified 44033edb

Update inline style code to multiline (#5291)

wonjoo-wj committed 2 years ago

Verified be8718ce

avoid copy proto in PrepareOutputShardingPropagation (#5287)

baoleai committed 2 years ago

Verified 54e2bb69

Lower squeeze.dims (#5286)

wonjoo-wj committed 2 years ago

Verified 865abd74

[Traceable Collecive] Hide token for all_gather (#5232)

alanwaketan committed 2 years ago

Verified 42d92709

Disable torch._dynamo.config.automatic_dynamic_shapes (#5285)

wonjoo-wj committed 2 years ago

Verified ebf9607b

Fix the error where mark_step does not materalize tensors on SPMD:0 (#5281)

JackCaoG committed 2 years ago

Verified cb5944ae

Make it an error if calling sizes() on a dynamic tensor. (#4998)

vanbasten23 committed 2 years ago

Verified d3779cb8

Add `runtime.host_index` (#5283)

will-cromar committed 2 years ago

Verified 68d577ee

Remove or improve several hardcoded TPU test conditions (#5272)

will-cromar committed 2 years ago

Verified b8617f52

Always do build_ext in python setup.py develop (#5273)

qihqi committed 2 years ago

Verified 71e6e9b0

Dsiable test_simple_model_with_different_input_shape since it is curretnly broken by pytorch (#5282)

JackCaoG committed 2 years ago

Verified b799e206

Newer Older