enable mixed precission

[divamgupta]

Thanks for the PR. It was already there in https://github.com/divamgupta/stable-diffusion-tensorflow/pull/8