pytorch/xla

Pull Requests Commits

Metrics comparator (#1402)

zcain117 committed 5 years ago

Verified 438f8429

Added debug_run.py section is docs and done some tweaks on it.

dlibenzi committed 5 years ago

ae9f5402

Allow metrics percentiles to be configurable.

dlibenzi committed 5 years ago

72588056

Do not add extra .0 if running single core. (#1437)

dlibenzi committed 5 years ago

b5ac7b12

Enable some dynamic shapes tests. (#1434)

dlibenzi committed 5 years ago

f66bf943

Properly namespace TorchGather(). (#1436)

dlibenzi committed 5 years ago

66fa4cd1

Fix single-core multi-processing.

dlibenzi committed 5 years ago

59ce59c7

Skip test on two logical ops

xuhdev committed 5 years ago

08cbd7af

Added debug script to run pytorch/xla models in supervised mode.

dlibenzi committed 5 years ago

77cce413

Allow parsing all graph types, not only TEXT.

dlibenzi committed 5 years ago

0a2728f3

Fix one-hot lowering. (#1425)

dlibenzi committed 5 years ago

d39fe2a8

Re-enable op-by-op test. (#1423)

dlibenzi committed 5 years ago

842b3abe

Bring new TF with AllReduce tuple fix.

dlibenzi committed 5 years ago

7ffd4d5e

Fake Token using scalar until XLA AllReduce() Token support is added.

dlibenzi committed 5 years ago

ec28190d

Make trim graph locations deterministic.

dlibenzi committed 5 years ago

8d2ae7c9

Fix the test by avoiding the current (will be fixed soon) XLA AllReduce bug on tuple inputs with single element inside.

dlibenzi committed 5 years ago

ea39a80a

Fix MP AllReduce test.

dlibenzi committed 5 years ago

93f6d209

Do not spawn a different process if we will be running on one single core.

dlibenzi committed 5 years ago

04944651

Allow pow() to take int exponents as the lowering does the promotions already.

dlibenzi committed 5 years ago

6e436eb4

Temporarily disable op-by-op test.

dlibenzi committed 5 years ago

0bd9a898

Use shape_fn for expensive shape calculations into GenericOp(). (#1410)

dlibenzi committed 5 years ago

954c99a2

Added aten::reflection_pad2d and aten::reflection_pad2d_backwar… (#1409)

dlibenzi committed 5 years ago

012367f6

Propagate requires_grad when sending to XLA device. (#1411)

dlibenzi committed 5 years ago

b618a917

Don't use auto in test

vishwakftw committed 5 years ago

c873a6dd

Remove logdet.h/logdet.cpp, and migrate implementation to ops.h/ops.cpp

vishwakftw committed 5 years ago

ecac1565

Fix build by making node

vishwakftw committed 5 years ago

9025c22a

Add aten::logdet op

vishwakftw committed 5 years ago

70c1ceaa

Migrate code to use the new XlaBuilder::GetShapePtr().

dlibenzi committed 5 years ago

b255f1d1

add hypothesis to conda install packages

zcain117 committed 5 years ago

078bd7b8

Bring new TF with GetShapePtr() changes.

dlibenzi committed 5 years ago

8ffcb772

Newer Older