Add `DWConvClass()` (#4274)
Browse files* Add `DWConvClass()`
* Cleanup
* Cleanup2
- models/common.py +9 -2
- models/experimental.py +1 -1
- models/yolo.py +2 -2
models/common.py
CHANGED
@@ -30,7 +30,7 @@ def autopad(k, p=None): # kernel, padding
|
|
30 |
|
31 |
|
32 |
def DWConv(c1, c2, k=1, s=1, act=True):
|
33 |
-
# Depth-wise convolution
|
34 |
return Conv(c1, c2, k, s, g=math.gcd(c1, c2), act=act)
|
35 |
|
36 |
|
@@ -45,10 +45,17 @@ class Conv(nn.Module):
|
|
45 |
def forward(self, x):
|
46 |
return self.act(self.bn(self.conv(x)))
|
47 |
|
48 |
-
def
|
49 |
return self.act(self.conv(x))
|
50 |
|
51 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
52 |
class TransformerLayer(nn.Module):
|
53 |
# Transformer layer https://arxiv.org/abs/2010.11929 (LayerNorm layers removed for better performance)
|
54 |
def __init__(self, c, num_heads):
|
|
|
30 |
|
31 |
|
32 |
def DWConv(c1, c2, k=1, s=1, act=True):
|
33 |
+
# Depth-wise convolution function
|
34 |
return Conv(c1, c2, k, s, g=math.gcd(c1, c2), act=act)
|
35 |
|
36 |
|
|
|
45 |
def forward(self, x):
|
46 |
return self.act(self.bn(self.conv(x)))
|
47 |
|
48 |
+
def forward_fuse(self, x):
|
49 |
return self.act(self.conv(x))
|
50 |
|
51 |
|
52 |
+
class DWConvClass(Conv):
|
53 |
+
# Depth-wise convolution class
|
54 |
+
def __init__(self, c1, c2, k=1, s=1, act=True): # ch_in, ch_out, kernel, stride, padding, groups
|
55 |
+
super().__init__(c1, c2, k, s, act)
|
56 |
+
self.conv = nn.Conv2d(c1, c2, k, s, autopad(k), groups=math.gcd(c1, c2), bias=False)
|
57 |
+
|
58 |
+
|
59 |
class TransformerLayer(nn.Module):
|
60 |
# Transformer layer https://arxiv.org/abs/2010.11929 (LayerNorm layers removed for better performance)
|
61 |
def __init__(self, c, num_heads):
|
models/experimental.py
CHANGED
@@ -72,7 +72,7 @@ class GhostBottleneck(nn.Module):
|
|
72 |
|
73 |
|
74 |
class MixConv2d(nn.Module):
|
75 |
-
# Mixed
|
76 |
def __init__(self, c1, c2, k=(1, 3), s=1, equal_ch=True):
|
77 |
super().__init__()
|
78 |
groups = len(k)
|
|
|
72 |
|
73 |
|
74 |
class MixConv2d(nn.Module):
|
75 |
+
# Mixed Depth-wise Conv https://arxiv.org/abs/1907.09595
|
76 |
def __init__(self, c1, c2, k=(1, 3), s=1, equal_ch=True):
|
77 |
super().__init__()
|
78 |
groups = len(k)
|
models/yolo.py
CHANGED
@@ -202,10 +202,10 @@ class Model(nn.Module):
|
|
202 |
def fuse(self): # fuse model Conv2d() + BatchNorm2d() layers
|
203 |
LOGGER.info('Fusing layers... ')
|
204 |
for m in self.model.modules():
|
205 |
-
if
|
206 |
m.conv = fuse_conv_and_bn(m.conv, m.bn) # update conv
|
207 |
delattr(m, 'bn') # remove batchnorm
|
208 |
-
m.forward = m.
|
209 |
self.info()
|
210 |
return self
|
211 |
|
|
|
202 |
def fuse(self): # fuse model Conv2d() + BatchNorm2d() layers
|
203 |
LOGGER.info('Fusing layers... ')
|
204 |
for m in self.model.modules():
|
205 |
+
if isinstance(m, (Conv, DWConvClass)) and hasattr(m, 'bn'):
|
206 |
m.conv = fuse_conv_and_bn(m.conv, m.bn) # update conv
|
207 |
delattr(m, 'bn') # remove batchnorm
|
208 |
+
m.forward = m.forward_fuse # update forward
|
209 |
self.info()
|
210 |
return self
|
211 |
|