1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
|
import torch
class Linear(torch.ao.nn.qat.Linear):
r"""
A linear module attached with FakeQuantize modules for weight,
used for dynamic quantization aware training.
We adopt the same interface as `torch.nn.Linear`, please see
https://pytorch.org/docs/stable/nn.html#torch.nn.Linear
for documentation.
Similar to `torch.nn.Linear`, with FakeQuantize modules initialized to
default.
"""
def __init__(self, in_features, out_features, bias=True,
qconfig=None, device=None, dtype=None) -> None:
super().__init__(in_features, out_features, bias, qconfig, device, dtype)
if not torch.ao.quantization.activation_is_memoryless(qconfig):
raise ValueError(
"Dynamic QAT requires a memoryless observer." +
"This means a MovingAverage observer with averaging constant equal to 1"
)
|
