ANSHUMAN87 opened a new issue #6791:
URL: https://github.com/apache/incubator-tvm/issues/6791
In Tensorflow version > 2.1.0, there is a bug induced in Tensorflow, where
if we try to use Quantized PadV2 with constant value input, it fails to covert
to TFLite because of mismatch in output quant params & the constant value quant
params.
Pseudo code:
` quant_output = pad_v2(quant_data, padding, mode="CONSTANT",
quant_constant_values)`
Error:
` RuntimeError: tensorflow/lite/kernels/pad.cc:205
op_context.output->params.zero_point !=
op_context.constant_values->params.zero_point (128 != 0)`
For test case, please
[refer](https://github.com/apache/incubator-tvm/blob/c94623ad0dac5fa5ce7a9a3c4ecb794351ebc610/tests/python/frontend/tflite/test_forward.py#L2561)
This is a tracker until bug is fixed in Tensorflow.
----------------------------------------------------------------
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]
