commit | 3e3503fd7157da1b7076056e6055971408d5c21e | [log] [tgz] |
---|---|---|
author | Jian Li <jianlijianli@google.com> | Mon Oct 21 15:01:59 2019 -0700 |
committer | TensorFlower Gardener <gardener@tensorflow.org> | Mon Oct 21 16:08:23 2019 -0700 |
tree | 0689f058bcadcf4049d8f34e68c7a28bdda3e4af | |
parent | 5f7e80591625672305e355c9c7cfbec146b1b34d [diff] |
Make "restrict_same_input_output_scale" field more clear. This is to prepare the operator properties for supporting fused ops. PiperOrigin-RevId: 275931432 Change-Id: I42ac9d95a0b18c808e86b01e29d250f2d1a6ded9
diff --git a/tensorflow/lite/tools/optimize/operator_property.h b/tensorflow/lite/tools/optimize/operator_property.h index b3c9975..39eeb60 100644 --- a/tensorflow/lite/tools/optimize/operator_property.h +++ b/tensorflow/lite/tools/optimize/operator_property.h
@@ -50,7 +50,7 @@ // Bias indexes. std::vector<int> biases = {}; - // Constraints. + // Force output to reuse the same scale and zero point of input. bool restrict_same_input_output_scale = false; // Op version.