Method quant_input

gpu/model.py:69–71

(self, input)

    @torch.compile
    def quant_input(self, input):
        s = 127 / input.abs().max(dim=-1, keepdim=True).values.clamp_(min=1e-5)
        return (input * s).round().clamp(-128, 127).to(torch.int8), s

    def forward(self, input):
        input, s = self.quant_input(input)