pytorch/xla

Pull Requests Commits

wonjoo-wj committed 3 years ago

d4f238b3

Rebase with master and clean up code

wonjoo-wj committed 3 years ago

600ff6b8

Full codegen addcmul

wonjoo-wj committed 3 years ago

92e3708c

Check if XLA_SANDBOX_BUIDL is unset before applying the bazel patch (#3590)

yeounoh committed 3 years ago

Verified 9a5cfac8

Update libtpu pin to 20220518 nightly (#3586)

yeounoh committed 3 years ago

Verified 8ca18441

Address review comments from #3550 (#3578)

will-cromar committed 3 years ago

Verified 7b3b598f

Remove existing torch_xla::OpList (#3583)

wonjoo-wj committed 3 years ago

Verified 3d24d955

Bazel build issue patch (#3570)

yeounoh committed 3 years ago

Verified bc6471f0

Pin all layout by default and use all_reduce based all_gather when la… (#3568)

JackCaoG committed 3 years ago

Verified 7f96fbf6

Replace XlaValue with upstream lazy::Value (#3572)

wonjoo-wj committed 3 years ago

Verified ec8a7447

Use upstream torch::lazy::Computation (#3580)

JackCaoG committed 3 years ago

Verified 82331cf1

Full codegen for `sgn`, `sign` (#3577)

miladm committed 3 years ago

Verified 041ebf92

codegen for cos and cosh (#3574)

miladm committed 3 years ago

Verified 0bef61c5

Experimental support for single-host TPU in PjRt (#3550)

will-cromar committed 3 years ago

Verified 1ab84187

Revert PR#3450 and use sparse_gather in gather (#3566)

yeounoh committed 3 years ago

Verified 90a525b5

Full codegen asin, asinh, atan, and atanh (#3565)

wonjoo-wj committed 3 years ago

Verified 56a52d53

Codegen acos and acosh ops (#3564)

wonjoo-wj committed 3 years ago

Verified 2cc13c08

Add send/recv support for https://github.com/pytorch/xla/issues/3138 (#3494)

hjm-aws committed 3 years ago

Verified 137415b7

Codegen Abs and Maximum op (#3544)

JackCaoG committed 3 years ago

Verified 7a306610

Shift int only (#3561)

JackCaoG committed 3 years ago

Verified 344c1b1d

Update for consolidated contiguous/sizes policy (#3555)

ezyang committed 3 years ago

Verified bc310a94

Implement Fully Sharded Data Parallel (FSDP) in PyTorch XLA (#3431)

ronghanghu committed 3 years ago

Verified 3c83269f

allow disabling layout pinning in optimizer_step and reduce_gradients (#3556)

ronghanghu committed 3 years ago

Verified 6e3992ac

Lower LogSigmoid, LogSigmoidBackward, Elu, and EluBackward (#3539)

wonjoo-wj committed 3 years ago

Verified cc312e27

Delete .torch_pin

JackCaoG committed 3 years ago

Verified 8baddaba

Remove unpool backward (#3537)

JackCaoG committed 3 years ago

Verified 87fda0ac

Update matmul cpp test to check a different counter (#3154)

JackCaoG committed 3 years ago

Verified aa34578e

Lower celu, celu_, selu, and selu_ (#3547)

wonjoo-wj committed 3 years ago

Verified d1de2477

Add ir_builder include since MakeNode moved to this file (#3548)

JackCaoG committed 3 years ago

Verified a3031d64

Start implementation of PjRtComputationClient. (#3463)

will-cromar committed 3 years ago

Verified 9525a8ec

Older