tree c5f881d9c02aa78fffffe51e74783d721520f57f
parent 5ef6caf543d0ab18d6df61c07e93e3a518d89add
author A. Unique TensorFlower <gardener@tensorflow.org> 1623623797 -0700
committer TensorFlower Gardener <gardener@tensorflow.org> 1623624194 -0700

Change tflite::PopulateConvolutionQuantizationParams() per_channel_shift parameter to be int32_t* instead of int*.

This brings it into line with TfLite Micro, which allocates int32_t storage for per_channel_shift parameters. See https://github.com/tensorflow/tflite-micro/issues/164.

PiperOrigin-RevId: 379185579
Change-Id: I349b3c7a7c7b29d768f649b3ab86b657d73402f6