This is an automated email from the ASF dual-hosted git repository. masahi pushed a commit to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/main by this push: new fb87c21bf8 remove duplicated cast op when lowering qnn.requantize op in float mode (#12234) fb87c21bf8 is described below commit fb87c21bf8d0fa5edec96a054a57a6d37c11289f Author: paperplanet <etb2...@gmail.com> AuthorDate: Sat Jul 30 16:28:39 2022 +0800 remove duplicated cast op when lowering qnn.requantize op in float mode (#12234) --- src/relay/qnn/op/requantize.cc | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) diff --git a/src/relay/qnn/op/requantize.cc b/src/relay/qnn/op/requantize.cc index 2a6153e810..5bf53a95ed 100644 --- a/src/relay/qnn/op/requantize.cc +++ b/src/relay/qnn/op/requantize.cc @@ -303,10 +303,7 @@ Expr RequantizeLowerFP(const Expr& input_tensor, const Expr& input_scale, -1, }), rank, {axis}); - tensor = Subtract(Cast(tensor, DataType::Float(Bits)), - Cast(input_zero_broadcast, DataType::Float(Bits))); - } else { - tensor = Cast(tensor, DataType::Float(Bits)); + tensor = Subtract(tensor, Cast(input_zero_broadcast, DataType::Float(Bits))); } // 2) If the input and output scales are same, we can skip the multiplication. Check