diff options
author | Adrian Kuegel <akuegel@google.com> | 2018-09-03 04:52:24 -0700 |
---|---|---|
committer | TensorFlower Gardener <gardener@tensorflow.org> | 2018-09-03 04:57:16 -0700 |
commit | 0f4ad7ff0e5ce38bc09ddd008e4f32d2af321495 (patch) | |
tree | cc76401cd5057d454fc0dc82178c3747a122cb3f /tensorflow/compiler/xla/service/hlo_instruction.h | |
parent | 44c884dc5d02abc7c50abea24c8caee6dcadda9a (diff) |
Call Cudnn also for grouped convolutions.
Cudnn supports grouped convolutions, so we don't need the
ConvolutionFeatureGroupConverter pass and can instead set the group_count
parameter on the cudnn custom calls.
PiperOrigin-RevId: 211339551
Diffstat (limited to 'tensorflow/compiler/xla/service/hlo_instruction.h')
-rw-r--r-- | tensorflow/compiler/xla/service/hlo_instruction.h | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/tensorflow/compiler/xla/service/hlo_instruction.h b/tensorflow/compiler/xla/service/hlo_instruction.h index 08f3d5356f..cca134e8b4 100644 --- a/tensorflow/compiler/xla/service/hlo_instruction.h +++ b/tensorflow/compiler/xla/service/hlo_instruction.h @@ -1475,6 +1475,8 @@ class HloInstruction { // dimension and output feature dimension. int64 feature_group_count() const; + void set_feature_group_count(int64 feature_group_count); + // Delegates to HloSelectAndScatterInstruction::select. HloComputation* select() const; |