pytorch/xla

Pull Requests Commits

Merge remote-tracking branch 'origin' into benchmark

zpcore committed 1 year ago

56e57f0f

Add hide-errors flag to result analyzer (#5836)

frgossen committed 1 year ago

a6bef634

Add flag to forward XLA flags and allow for experiment expansion (#5828)

frgossen committed 1 year ago

cea3ef1d

Fix syntax in experiment_runner.py (#5827)

frgossen committed 1 year ago

ac1b5324

benchmarks: add script to aggregate results from result_analyzer (#5829)

cota committed 1 year ago

9b92890f

Add profiling capabilities to experiment_runner.py script (#5812)

golechwierowicz committed 1 year ago

74cb8db4

Add top tier benchmark runner (#5809)

frgossen committed 1 year ago

d47a8e60

Apply Pytorch/XLA formatting style (#5816)

frgossen committed 1 year ago

4d02b650

Add benchmark selection by tier 1-3 (#5808)

frgossen committed 1 year ago

a0420873

Add python 3.9 and 3.11 build for nightly (#5840)

JackCaoG committed 1 year ago

Verified 57e60352

Add a unittest testing each core ATen ops. (#5796)

qihqi committed 1 year ago

Verified 2b04794a

Sync `xla_args` before computation. (#5823)

ysiraichi committed 1 year ago

Verified f80e3ba9

Use cuda 12.1 and build pytorch with cuda enabled. (#5831)

vanbasten23 committed 1 year ago

Verified 8c03ac96

Configure PATH and LD_LIBRARY_PATH to include cuda (#5761)

vanbasten23 committed 1 year ago

Verified 4fb54281

Extend HLO metadata to include class hierarchy information (#5715)

mrnikwaws committed 1 year ago

Verified 5e503794

update optimizer for openxla

zpcore committed 1 year ago

5c65c7a9

[SPMD] fix bug with XLAShardedTensor.__repr__ (#5807)

yeounoh committed 1 year ago

Verified 6addbde3

Fix device names and dynamo backend names in benchmark runner (#5806)

frgossen committed 1 year ago

Verified 7473cc45

Use upstream XLA concurrency utilities (#5799)

will-cromar committed 1 year ago

Verified 6e661304

Add python 3.11 trigger for 2.1 release build (#5810)

JackCaoG committed 1 year ago

Verified a3bbbb1e

Enable passing down dynamic dimensions from torch to XLA (#5790)

lsy323 committed 1 year ago

Verified a0a048e1

Don't set $TPU_LIBRARY_PATH during import

will-cromar committed 1 year ago

Verified d8c6a83f

Record the lazy tracing time(C++) in metrics (#5757)

JackCaoG committed 1 year ago

Verified 58ce7e17

Use TPU profiler plugin (#5793)

will-cromar committed 1 year ago

Verified 3b34cb27

Add GKE support and various usability improvements in CheckpointManager (#5770)

jonb377 committed 1 year ago

Verified 5710a83c

Lower AtenFull op (#5781)

danielvegamyhre committed 1 year ago

Verified b3ed82c9

Transfer data directly to the device (#5772)

will-cromar committed 1 year ago

Verified 05a3cdda

[SPMD] suppor DTensor API integration (#5776)

yeounoh committed 1 year ago

Verified e60428d1

disable test_set temporarily (#5792)

yeounoh committed 1 year ago

Verified 6f4183c8

Rename run_benchmark to run_benchmark.sh

ManfeiBai committed 1 year ago

Verified ee5acc1f

Older