Yeah, but it’s ’quantization aware’ during training too, which presumably is what allows the quantization at inference to work
Yeah, but it’s ’quantization aware’ during training too, which presumably is what allows the quantization at inference to work