Pytorch support int16 quantization

Hello, my friends

I have a question about int16 quantization of pytorch. Does pytorch have the plan to support int16 quantization in the future?  I'm looking forward to your reply. Thanks.

maybe, it depends on the trend in the hardwares we want to support, we do not have concrete plans to do that currently though. Please let us know if you have a prevailing use case for that.