pytorch/xla

Pull Requests Commits

Move 0-dimensional tensors to CPU before copying to XLA. (#6071)

ysiraichi committed 1 year ago

55279459

[Cherry-pick] [OpenXLA-pin] Update OpenXLA-pin to Nov29 (#6053) (#6079)

ManfeiBai committed 1 year ago

Verified 17cf6741

Add documentation and python API for persistent cache (#6074)

jonb377 committed 1 year ago

Verified a61ddef3

[Backport] Move CUDA tensors to CPU before moving to XLA (#6066)

will-cromar committed 1 year ago

Verified f77d1fac

[Backport]Add a spmd for application to pass in topology (#5931) (#6054)

JackCaoG committed 1 year ago

Verified 45742d2d

Enable persistent compilation caching (#6065)

jonb377 committed 1 year ago

Verified 01405e18

[Backport] Add reduce-scatter coalescing for FSDP/ZeRO1 (#5956) (#6055)

JackCaoG committed 1 year ago

Verified 8472abdf

[Cherry-Pick] Incorporate compilation environment in hash (#6050)

jonb377 committed 1 year ago

Verified b576d85f

[Backport][dynamo] Move CPU tensor constructor nodes to XLA. (#5922) (#6052)

JackCaoG committed 1 year ago

Verified aaccf540

Make PT_XLA_DEBUG analysis's number of frame printed configurable (#6037)

JackCaoG committed 1 year ago

Verified 616e91e6

Add a persistent cache (#5800)

jonb377 committed 1 year ago

Verified f06f6c00

Fix bug with IsTpuDevice for PJRT_DEVICE=TPU_C_API (#6031)

yeounoh committed 1 year ago

Verified 216e07dc

Enable Cumsum with long input (#5974)

JackCaoG committed 1 year ago

Verified 0287a3b5

Enable reduce scatter test on gpu (#6019)

JackCaoG committed 1 year ago

Verified b7cf37af

Disable flaky test (#6021)

JackCaoG committed 1 year ago

Verified 062aa91a

Update libtpu version to Nov24(#6026)

ManfeiBai committed 1 year ago

Verified e2229fd5

Update OpenXLA-pin to Nov24 (#6012)

ManfeiBai committed 1 year ago

Verified 729bae5e

[SPMD] Add debug of SPMD for single/multi host (#5742)

ManfeiBai committed 1 year ago

Verified f94b53cf

Fix test_dynamo_input_sharding_threashold (#5957)

JackCaoG committed 1 year ago

Verified 513be64b

fix docker issue fore xla GPU run on H100 (#6018)

zpcore committed 1 year ago

Verified 75d94fee

Add profiler API for async capture (#5969)

jonb377 committed 1 year ago

Verified 6d73ca8e

Add graph hash to save tensor output (#5971)

JackCaoG committed 1 year ago

Verified 495c820c

Make the fsdp test run by torchrun (#5916)

vanbasten23 committed 1 year ago

Verified f80a4149

[benchmarks] trivial: fix typo (#6016)

cota committed 1 year ago

Verified 35c53332

[Export] Add an option to not save the weights (#5964)

ManfeiBai committed 1 year ago

Verified 77b968a5

Fix duplicate @unittest.skip test_core_aten_ops.py (#5976)

wonjoo-wj committed 1 year ago

Verified 990c673c

Add script for updating core aten opset issue (#5821)

wonjoo-wj committed 1 year ago

Verified 74fd6e5f

Optimize autocast tests. (#5970)

vanbasten23 committed 1 year ago

Verified 67cb9ca6

Allow benchmark runner config: `openxla` for inference runs. (#5939)

ysiraichi committed 1 year ago

Verified dcdd66e3

Add all-gather coalescing for FSDP/ZeRO1 (#5950)

jeffhataws committed 1 year ago

Verified 12719646

Older