pytorch/xla

Pull Requests Commits

alanwaketan committed 1 year ago

faeb7c79

Disallow mark manual sharding on data tensors

alanwaketan committed 1 year ago

84cd609b

Fix one comment

alanwaketan committed 1 year ago

2db7f0a9

alanwaketan committed 1 year ago

995af75b

alanwaketan committed 1 year ago

e666e3b5

alanwaketan committed 1 year ago

39008eb0

alanwaketan committed 1 year ago

d77e011c

Update XLA pin to 04/08/2024 (#6904)

JackCaoG committed 1 year ago

Verified fd04a20c

[torch_xla2] Implement some common tensor constructors (#6874)

will-cromar committed 1 year ago

Verified 786484d2

[Pallas] Remove torch.empty in tracing (#6897)

alanwaketan committed 1 year ago

Verified 66ed39ba

Set all_reduce_token to null when exiting (#6898)

yitongh committed 1 year ago

Verified bd2b8fc8

update rc12 for 2.3 release (#6903)

lsy323 committed 1 year ago

Verified 1fe6be30

Add dynamo `torch.Tensor.new` test. (#6661)

ysiraichi committed 1 year ago

Verified 139d891e

Fix README and add partner links (#6896)

JackCaoG committed 1 year ago

Verified 5c48be19

add torch_distributed tests to our TPU CI (#6891)

JackCaoG committed 1 year ago

Verified 44bf47c4

Bump JAX version to match the latest pin update (#6893)

alanwaketan committed 1 year ago

Verified 77be46bd

[SPMD] UseVirtualDevice() is set, when XLA_AUTO_SPMD=1 (#6894)

yeounoh committed 1 year ago

Verified f44875c0

Manually init absl log to avoid log spam (#6890)

JackCaoG committed 1 year ago

Verified 62f6e05c

use init_method=xla:// for torch.distributed test (#6888)

JackCaoG committed 1 year ago

Verified 194ac22a

Update xla pin to Apr 3 (#6879)

bhavya01 committed 1 year ago

Verified 254970e5

Add opinfo test (#6878)

qihqi committed 1 year ago

Verified 0b9513cb

Adjust allclose atol for the flash attention TPU test (#6889)

JackCaoG committed 1 year ago

Verified 73de972b

Don't bundle libtpu in nightly (#6885)

alanwaketan committed 1 year ago

Verified a8f448d1

[Pallas] Make FlashAttention as torch.autograd.Function (#6886)

alanwaketan committed 1 year ago

Verified 0c704cf8

[benchmarks] Use Torchbench YAML configuration data. (#6881)

ysiraichi committed 1 year ago

Verified 7520c046

Fix vector norm lowering (#6883)

lsy323 committed 1 year ago

Verified f4371b67

Add flash attention as custom op so dynamo can trace it (#6875)

JackCaoG committed 1 year ago

Verified d9386809

Check pointer when we cast PjRtLayout to PjRtXlaLayout (#6882)

vanbasten23 committed 1 year ago

Verified 17823c3b

Fix div lowering and core aten test script enhancement (#6873)

lsy323 committed 1 year ago

Verified 895b0c2e

[Pallas] Support Flash Attention backward kernels (#6870)

alanwaketan committed 1 year ago

Verified c54367c8

Older