pytorch/xla

Pull Requests Commits

remove the usage of devkind

JackCaoG committed 1 year ago

6781fd40

[Backport] Remove reshape lowering (#6801) (#6847)

wonjoo-wj committed 1 year ago

Verified 6e4b2ed4

[r2.3] Fix div lowering and core aten test script enhancement (#6873) (#6877)

lsy323 committed 1 year ago

Verified fabf9778

[Backport] Set the major-minor layout for inputs and outputs (#6876)

alanwaketan committed 1 year ago

Verified 3905372e

[r2,3][SPMD] use UNKNOWN as default sharded data sharding type selectively (#6852)

yeounoh committed 1 year ago

Verified c3d57654

[r2.3] Run pre-submit, post-submit TPU CI on the relesae branch (#6856) (#6859)

lsy323 committed 1 year ago

Verified 0d1f1723

[r2.3] Use AddBufferDonor to replace SetUpAlias for input output aliasing (#6855)

lsy323 committed 1 year ago

Verified 13a2169f

udpate release tag to rc6 (#6839)

lsy323 committed 1 year ago

Verified 235864a0

Make r2.3 to be compatible with PyPI (#6838)

lsy323 committed 1 year ago

Verified 1baea32d

[Backport] Support XLA_USE_BF16 (#6841)

alanwaketan committed 1 year ago

Verified c6a88740

[Backport] Allow setting FlashAttention's causal mask (#6837)

alanwaketan committed 1 year ago

Verified 046f9103

Document the CUDA plugin in r2.3 README (#6825)

will-cromar committed 1 year ago

Verified 8de8cea3

[Backport] Make a FlashAttention Wrapper (#6827)

alanwaketan committed 1 year ago

Verified db7112af

Backport ZeRO1: Add bucketting logic to control the size of tensors for all-gather/reduce-scatter (#6025) (#6806)

JackCaoG committed 1 year ago

Verified a805505d

[r2.3] Update XLA pin (#6797) (#6811)

yeounoh committed 1 year ago

Verified 7e8b8fb7

[Pallas] Add a bfloat16 flash attention test case (#6810)

alanwaketan committed 1 year ago

Verified 3522be16

Properly lower div.scalar and div.tensor(#6669) (#6808)

bhavya01 committed 1 year ago

Verified 8240d05b

Backport Fix preserve_rng_state for activation checkpointing (#4690) (#6788)

JackCaoG committed 1 year ago

Verified 1d6558a3

[Backport] Introduce jax_import_guard (#6794)

alanwaketan committed 1 year ago

Verified c3001d64

Address deprecation in torch.distributed.checkpoint (#6786)

jonb377 committed 1 year ago

Verified d9877757

[SPMD] auto-construct auto-sharding mesh ids (#6770) (#6782)

yeounoh committed 1 year ago

Verified c923e8f1

backport `remove openxla_eval from the doc` (#6767)

JackCaoG committed 1 year ago

Verified ebedf4d9

[r2.3] Include auto-sharding configs as part of hash (#6753) (#6763)

yeounoh committed 1 year ago

Verified a78586f8

[r2.3] backport: auto-sharding PoC (#6719) (#6755)

yeounoh committed 1 year ago

Verified 680ebc10

Properly lower add and mul (#6731) (#6744)

lsy323 committed 1 year ago

Verified f9c94d9a

Properly lower aten::sub and aten::rsub (#6668) (#6743)

lsy323 committed 1 year ago

Verified 9cb18dfa

[Pallas] Introduce make_kernel_from_pallas (#6713) (#6742)

lsy323 committed 1 year ago

Verified e556ad81

Backport Bump JAX version in `setup.py` (#6740) (#6741)

lsy323 committed 1 year ago

Verified a47f81d9

Update `torch_xla_cuda_plugin` version for r2.3 (#6729)

will-cromar committed 1 year ago

Verified aa6d2775

add torch pin to 2.3 release branch (#6735)

lsy323 committed 1 year ago

Verified 45268bd1

Older