pytorch/xla

Pull Requests Commits

Make PT_XLA_DEBUG analysis's number of frame printed configurable (#5958)

JackCaoG committed 1 year ago

b25bd652

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

Increase tolerance for tan (#5915)

qihqi committed 1 year ago

Verified 2c4983dd

Enable TF32 multiplication by default. (#5965)

golechwierowicz committed 1 year ago

Verified e98cb665

Vectorize local shard retrieval (#5826)

jonb377 committed 1 year ago

Verified c919973d

use a common random seed for test core aten ops (#5949)

qihqi committed 1 year ago

Verified f6a775c4

Distribute Literal->Tensor copies across thread pool (#5825)

jonb377 committed 1 year ago

Verified ec54fd44

Add stack frame id metadata and correct filename and line number for custom op_name (#5838)

mrnikwaws committed 1 year ago

Verified 5fa89223

Update troubleshooting doc with few common env var combination (#5929)

JackCaoG committed 1 year ago

Verified 6c5b7b8e

Add graph hash and num input/output to PT_XLA_DEBUG (#5947)

JackCaoG committed 1 year ago

Verified f3b75ba1

Refactor ExecuteReplicated to operate on sharded data directly (#5737)

will-cromar committed 1 year ago

Verified 7b92a949

Older