quantized-nets
quantized-nets copied to clipboard
Are training and evaluations both Quantized?
In the implementation, it looks like training is Quantized form but evaluation/test is still in Full precision. Is this assumption correct?