pytorch/xla

Pull Requests Commits

use dim > 2 to suit TPU v3(otherwise can't be divide evenly)

JackCaoG committed 2 years ago

3affe2a4

use permute instead of transpose

JackCaoG committed 2 years ago

9f92496a

Supoort non-ordered sharding spec correctly

JackCaoG committed 2 years ago

6b8b8ece

[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

Fix the error when export_torch_model is given a non-tensor (#5277)

qihqi committed 2 years ago

Verified 23e722f6

Delete unused .so file, and .lds files (#5275)

ManfeiBai committed 2 years ago

Verified ba01cf9a

Clean bazel stuff on distutils clean. (#5274)

stgpetrovic committed 2 years ago

Verified 1842414c

Make TPU detection more robust (#5271)

will-cromar committed 2 years ago

Verified 048f3fa6

[SPMD] Sharding n-d tensor on (n+1)-d Mesh (#5268)

khatwanimohit committed 2 years ago

Verified adc97adc

Suppress debug symbols in OpenXLA code (#5269)

stgpetrovic committed 2 years ago

Verified 625d2d07

Disable Bazel remote cache for forked PR (#5259)

lsy323 committed 2 years ago

Verified 81c5cc07

Enable PjRt Client Compilation with StableHLO (#5233)

lsy323 committed 2 years ago

Verified 905be9d5

Older