As per the documentation, PyTorch will support int8 quantization. Will PyTorch support int16 quantization currently?
We currently do not support int16 quantization. There is support for fp16 dynamic quantization.
1 Like