masahi commented on a change in pull request #7474:
URL: https://github.com/apache/tvm/pull/7474#discussion_r578915463
##########
File path: src/relay/qnn/op/dequantize.cc
##########
@@ -105,6 +115,10 @@ Expr DequantizeLower(const Expr& input_tensor, const Expr&
input_scale,
auto shift = Subtract(Cast(input_tensor, DataType::Int(32)),
expanded_input_zero_point);
auto scaled_output = Multiply(Cast(shift, DataType::Float(32)),
expanded_input_scale);
+
+ if (out_dtype != DataType::Float(32)) {
+ scaled_output = Cast(scaled_output, out_dtype);
Review comment:
Casting the output of float multiply to int is wrong.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
