|
|
import torch |
|
|
import torch.nn as nn |
|
|
|
|
|
|
|
|
def w8_a16_forward(weight, input, scales, bias=None): |
|
|
|
|
|
casted_weights = weight.to(input.dtype) |
|
|
output = F.linear(input, casted_weights) * scales |
|
|
|
|
|
if bias is not None: |
|
|
output = output + bias |
|
|
|
|
|
return output |
|
|
|
|
|
class W8A16LinearLayer(nn.Module): |
|
|
def __init__(self, in_features, out_features, |
|
|
bias=True, dtype=torch.float32): |
|
|
super().__init__() |
|
|
|
|
|
|
|
|
self.register_buffer( |
|
|
"int8_weights", |
|
|
torch.randint( |
|
|
-128, 127, (out_features, in_features), dtype=torch.int8 |
|
|
) |
|
|
) |
|
|
|
|
|
self.register_buffer("scales", |
|
|
torch.randn((out_features), dtype=dtype)) |
|
|
|
|
|
if bias: |
|
|
self.register_buffer("bias", |
|
|
torch.randn((1, out_features), |
|
|
dtype=dtype)) |
|
|
|
|
|
else: |
|
|
self.bias = None |
|
|
|
|
|
def quantize(self, weights): |
|
|
w_fp32 = weights.clone().to(torch.float32) |
|
|
|
|
|
scales = w_fp32.abs().max(dim=-1).values / 127 |
|
|
scales = scales.to(weights.dtype) |
|
|
|
|
|
int8_weights = torch.round(weights |
|
|
/scales.unsqueeze(1)).to(torch.int8) |
|
|
|
|
|
self.int8_weights = int8_weights |
|
|
self.scales = scales |
|
|
|
|
|
def forward(self, input): |
|
|
return w8_a16_forward(self.int8_weights, |
|
|
input, self.scales, self.bias) |
|
|
|
