u99127 commented on a change in pull request #6018:
URL: https://github.com/apache/incubator-tvm/pull/6018#discussion_r470636451
##########
File path: tests/python/frontend/tflite/test_forward.py
##########
@@ -250,7 +256,7 @@ def compare_tflite_with_tvm(in_data, in_name, input_tensors,
# convert to tflite model
converter = tf.lite.TFLiteConverter.from_session(
sess, input_tensors, output_tensors)
-
+ converter.experimental_new_converter=experimental_new_converter
Review comment:
@anijain2305 - sorry for some reason github is refusing to send
notifications to email when tagged here :( and that's the reason for my delay
in responding to this in addition to holidays.
AFAIUI there is no way of freezing a tflite model that contains quantized
max or min using the toco converter and thus we need to use the API in that
form to get the testsuite coverage. While the API to use this is "subject to
change", from my pov it's a use in the testsuite , we aren't using it in the
main code base and thus using it is less risky.
Note also that the tflite converter in tensorflow is now defaulting to the
mlir based converter as per the latest docs so this use is still a conservative
move forward as we are sticking to the default but it gives us additional
operator coverage.
Does that help ?
Ramana
----------------------------------------------------------------
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]
