75_Contribute_new_quantization_method
class Linear4bit(nn.Module): def __init__(self, ...): ... def forward(self, x): return my_4bit_kernel(x, self.weight, self.bias)