Quantization ideas

Low-precision training and inference formulations that keep accuracy within the benchmark's noise floor.