support q8_0 (#8765)
This commit is contained in:
parent
394304b918
commit
4afea496ab
1 changed files with 1 additions and 1 deletions
|
|
@ -247,7 +247,7 @@ class LinearQuant(nn.Linear):
|
||||||
if x_2d.shape[0] > 1 and x_2d.dtype == torch.float32:
|
if x_2d.shape[0] > 1 and x_2d.dtype == torch.float32:
|
||||||
x_2d = x_2d.half()
|
x_2d = x_2d.half()
|
||||||
# input format of linear_q4.forward is 1: input, 2: weight
|
# input format of linear_q4.forward is 1: input, 2: weight
|
||||||
result = linear_q4_0.forward(x_2d, x0)
|
result = linear_q4_0.forward(x_2d, x0, self.qtype)
|
||||||
new_shape = x_shape[:-1] + (self.out_len,)
|
new_shape = x_shape[:-1] + (self.out_len,)
|
||||||
result = result.view(new_shape)
|
result = result.view(new_shape)
|
||||||
if self.bias is not None:
|
if self.bias is not None:
|
||||||
|
|
|
||||||
Loading…
Reference in a new issue