pytorch/xla

Pull Requests Commits

Merge branch 'master' into fix_aten_prod

LukeBoyer committed 2 years ago

ff667dc1

Set all_reduce_token to None when exiting (#6247)

yitongh committed 2 years ago

Verified f9c12fc1

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

vanbasten23 committed 2 years ago

Verified 99cfeee6

Experimental GPU PJRT Plugin (#6240)

will-cromar committed 2 years ago

Verified 29977cbd

Remove obsolete device special cases (#6197)

will-cromar committed 2 years ago

Verified 0cd6f105

update release rc wheel name to RC6 (#6245)

ManfeiBai committed 2 years ago

Verified d96c15d6

Remove reshape lowering (#6230)

JackCaoG committed 2 years ago

Verified 0c81365e

Fix some core aten ops (#6243)

wonjoo-wj committed 2 years ago

Verified d7c44308

Create sharded data placeholders directly (#6203)

will-cromar committed 2 years ago

Verified efa6fcfd

Remove Colab notebooks (#6155)

will-cromar committed 2 years ago

Verified 26dcba94

LukeBoyer committed 2 years ago

3f1c3e16

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

cota committed 2 years ago

Verified 99fdf9a2

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

cota committed 2 years ago

Verified 9c320743

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

ManfeiBai committed 2 years ago

Verified 22a6e6ea

Fix erf, atan2 for integers (#6216)

qihqi committed 2 years ago

Verified aff50b80

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

cota committed 2 years ago

Verified bcefe806

Pin expecttest version (#6221)

JackCaoG committed 2 years ago

Verified 0f16471e

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

wonjoo-wj committed 2 years ago

Verified c9c2f3bd

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

wonjoo-wj committed 2 years ago

Verified 7339c82e

Add manual type conversion for aten_asin (#6213)

wonjoo-wj committed 2 years ago

Verified 8af15d04

Update dtype conversion for aten_tan (#6211)

wonjoo-wj committed 2 years ago

Verified 051a0be1

Enable some core aten tests (#6215)

qihqi committed 2 years ago

Verified e5039866

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

cota committed 2 years ago

Verified 86636ba7

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

cota committed 2 years ago

Verified dc8948d9

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

cota committed 2 years ago

Verified 0eb9afbe

benchmarks/torchbench_model: skip benchmarks that fail to load (#6199)

cota committed 2 years ago

Verified 3ccb4ed2

Dump dynamo counters (#6201)

frgossen committed 2 years ago

Verified 9a92472e

benchmarks/aggregate: make fig height and width configurable (#6209)

cota committed 2 years ago

Verified 56b72c41

Add option to dump HLOs during benchmark runs (#6200)

frgossen committed 2 years ago

Verified 61d42ebe

experiment_runner: revamp options to dump PyTorch/XLA metrics (#6186)

frgossen committed 2 years ago

Verified ad7c1412

Older