pytorch/xla

Pull Requests Commits

address comment

alanwaketan committed 1 year ago

1bca5aee

alanwaketan committed 1 year ago

98197faa

alanwaketan committed 1 year ago

042d2cd1

Remove XLA_TPU_LAYOUT

alanwaketan committed 1 year ago

c9d14d6f

alanwaketan committed 1 year ago

e5782cb3

Disable failed C++ Test (#6823)

lsy323 committed 1 year ago

Verified 2b66e93f

Add experimental MLIR debuginfo writer API (#6799)

chunnienc committed 1 year ago

Verified 64a146d8

Use fsspec for multihost checkpoint (#6818)

jonb377 committed 1 year ago

Verified 1ad6bb45

[Pallas] Support XLA_USE_BF16 (#6817)

alanwaketan committed 1 year ago

Verified 73c31dbe

Revert "Install torchvision in release docker" (#6821)

lsy323 committed 1 year ago

Verified b8a97dfc

Add missing shape metadata for the last replaced fx node (#6805)

lsy323 committed 1 year ago

Verified 899a0fa4

Revert "[Reland]Change ATEN generator argument type to const std::optional<Generator>&" (#6819)

alanwaketan committed 1 year ago

Verified 134f5b62

[Fori_loop|While_loop] Add nested test (#6807)

ManfeiBai committed 1 year ago

Verified 22fe1dc3

Install torchvision in release docker (#6795)

lsy323 committed 1 year ago

Verified 7b9177de

Remove reshape lowering (#6801)

wonjoo-wj committed 1 year ago

Verified a84509b2

[Reland]Change ATEN generator argument type to const std::optional<Generator>& (#6804)

cyyever committed 1 year ago

Verified 84e7feb3

Delete torch_patches/.torch_pin (#6812)

JackCaoG committed 1 year ago

Verified 42ecc295

Make compile count in Dynamo test more general (#6802)

JackCaoG committed 1 year ago

Verified 2e5c3b64

Inline all scalars by default in export path (#6803)

lsy323 committed 1 year ago

Verified f43ac41e

Add missing import for XLA init_method (#6809)

jonb377 committed 1 year ago

Verified 34d17c2d

Update XLA pin (#6797)

yeounoh committed 1 year ago

Verified b0ceb2ba

Add cuda 12.1 nightly wheels for cp310 and cp311. (#6793)

vanbasten23 committed 1 year ago

Verified 7d8972c8

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

jeffhataws committed 1 year ago

Verified e75677f1

Update the instruction of installing the NVIDIA container toolkit (#6791)

vanbasten23 committed 1 year ago

Verified 782f05d2

[Pallas] Allow setting FlashAttention's causal mask (#6792)

alanwaketan committed 1 year ago

Verified de42834d

Add patch for unbounded dynamism (#6783)

ghpvnist committed 1 year ago

Verified 8d579f91

Use stablehlo.composite instead of stablehlo.custom_call (#6789)

mlevesquedion committed 1 year ago

Verified 3eeb15d9

Use AddBufferDonor to replace SetUpAlias for input output aliasing (#6769)

JackCaoG committed 1 year ago

Verified d70b6c4e

Add dynamism support to `aten.embedding` and `aten.split_with_sizes` (#6781)

lsy323 committed 1 year ago

Verified 6b76c7f6

[Pallas] Make a FlashAttention Wrapper (#6785)

alanwaketan committed 1 year ago

Verified fcf24b6c

Older