TVM-QNN-concatenate-operator

CallNode is not considered when canonicalizing qnn.concatenate

The TVM commit da27e6d is tested.
In test_yolo.py, we dispatch qnn.concatenate to generate CallNode: run python3 test_yolo.py to reproduce the problem.
A fixed qnn.concatenate is provided. Please copy it to src/relay/qnn/op of TVM, re-build TVM, and then you can get input scale and zero point of CallNode.

Name		Name	Last commit message	Last commit date
Latest commit History 5 Commits
2.jpg		2.jpg
README.md		README.md
concatenate.cc		concatenate.cc
test_yolo.py		test_yolo.py
tuple_input_scales-failed.txt		tuple_input_scales-failed.txt
yolov3-416-int8_tf2_3_0.tflite		yolov3-416-int8_tf2_3_0.tflite