Is there support for per channel quantization when using mx.contrib.quant.quantize_model? Or some other way?
mx.contrib.quant.quantize_model