pytorch/xla

Pull Requests Commits

Measure GPU wall time without launch overheads. (#6233)

golechwierowicz committed 1 year ago

Verified bd82beb9

Set min TPU nodes to 0 for TPU CI & disable workflow run on push (#6270)

mbzomowski committed 1 year ago

Verified 257f0f53

test_zero1 passes on GPU with latest torch/xla (commit e1c94df) (#6268)

jeffhataws committed 1 year ago

Verified a60f8e7c

remove model training using openxla backend (#6262)

zpcore committed 1 year ago

Verified a89ebefd

Add workflow file for TPU CI (#6198)

mbzomowski committed 1 year ago

Verified 5801f7e5

Remove skip from prod lowering test (#6265)

LukeBoyer committed 1 year ago

Verified e1c94dfa

Remove skip from logical_or lowering test (#6264)

LukeBoyer committed 1 year ago

Verified 9eaab55a

Set all_reduce_token to None when exiting (#6247)

yitongh committed 1 year ago

Verified f9c12fc1

Set devices argument in dist.Backend.register_backend call. (#6254)

vanbasten23 committed 1 year ago

Verified 99cfeee6

Experimental GPU PJRT Plugin (#6240)

will-cromar committed 1 year ago

Verified 29977cbd

Remove obsolete device special cases (#6197)

will-cromar committed 1 year ago

Verified 0cd6f105

update release rc wheel name to RC6 (#6245)

ManfeiBai committed 1 year ago

Verified d96c15d6

Remove reshape lowering (#6230)

JackCaoG committed 1 year ago

Verified 0c81365e

Fix some core aten ops (#6243)

wonjoo-wj committed 1 year ago

Verified d7c44308

Create sharded data placeholders directly (#6203)

will-cromar committed 1 year ago

Verified efa6fcfd

Remove Colab notebooks (#6155)

will-cromar committed 1 year ago

Verified 26dcba94

benchmarks: add nightly.sh to automate regular benchmarking (#6231)

cota committed 1 year ago

Verified 99fdf9a2

experiment_runner: catch AMP exception when building child environment (#6229)

cota committed 1 year ago

Verified 9c320743

Promote int to float for acos operation (consistent with Pytorch) (#6224)

ManfeiBai committed 1 year ago

Verified 22a6e6ea

Fix erf, atan2 for integers (#6216)

qihqi committed 1 year ago

Verified aff50b80

Revert "Re-land: Fix model initialization. (#6182)" (#6220)

cota committed 1 year ago

Verified bcefe806

Pin expecttest version (#6221)

JackCaoG committed 1 year ago

Verified 0f16471e

Enable core aten test and loose tolerance for aten_log10 and aten_log2 (#6222)

wonjoo-wj committed 1 year ago

Verified c9c2f3bd

Update aten_exp to do a manual type conversion (#6214)

wonjoo-wj committed 1 year ago

Verified 7339c82e

Add manual type conversion for aten_asin (#6213)

wonjoo-wj committed 1 year ago

Verified 8af15d04

Update dtype conversion for aten_tan (#6211)

wonjoo-wj committed 1 year ago

Verified 051a0be1

Enable some core aten tests (#6215)

qihqi committed 1 year ago

Verified e5039866

benchmark/experiment_runner: save accelerator_model on error (#6218)

cota committed 1 year ago

Verified 86636ba7

benchmarks/aggregate: annotate latest plot with model names (#6217)

cota committed 1 year ago

Verified dc8948d9

benchmarks/aggregate: skip first three data points (#6210)

cota committed 1 year ago

Verified 0eb9afbe

Newer Older