pytorch/xla

Pull Requests Commits

version selector (#3178)

miaoshasha committed 2 years ago

ff037f14

Move FSDP README from Ronghang pr's comment (#3678) (#3679)

JackCaoG committed 3 years ago

Verified 883bfe9b

Include missing dependencies in setup.py (#3611) (#3650)

wonjoo-wj committed 3 years ago

Verified 82fbe570

Fix ScopePusher regression issue (#3637) (#3644)

wonjoo-wj committed 3 years ago

Verified 60d39ee6

More conservative IsSparseGather check to favor dense gather on TPU (#3630)

yeounoh committed 3 years ago

Verified eaba9288

Update sparsity check heuristics used in TorchGather (#3610) (#3613)

yeounoh committed 3 years ago

Verified 583731fa

Allowing `float16` and `bfloat16` computation type and sharding groups in XLA FSDP (#3588) (#3617)

wonjoo-wj committed 3 years ago

Verified 2dc32223

Hotfix cloudbuild bazel update (#3595)

yeounoh committed 3 years ago

64f32bab

Pin upstream pytorch for pytorch/xla 1.12 (#3603)

JackCaoG committed 3 years ago

Verified 94829463

updated googletest dependency to its stable version 1.11 pin (#3592)

miladm committed 3 years ago

Verified cc89a31d

Full codegen sin, sinh, and tan (#3581)

wonjoo-wj committed 3 years ago

Verified 352dd8c3

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

Older