AdaptiveAvgPool2D 不支持 onnx 导出，导出过程会告诉你，onnx不支持那个动态操作巴拉巴拉
我用的是 pp_liteseg 导出为 onnx 模型，都一样，paddle和Torch的 Adaptive Pool2D 都是动态的，onnx暂时都不支持，我根据下述公式，将 AdaptiveAvgPool2D 改了，可以用onnx导出了，但是需要指定一下原图的大小，和输出图的大小
h
s
t
a
r
t
=
f
l
o
o
r
(
i
∗
H
i
n
/
H
o
u
t
)
h
e
n
d
=
c
e
i
l
(
(
i
+
1
)
∗
H
i
n
/
H
o
u
t
)
w
s
t
a
r
t
=
f
l
o
o
r
(
j
∗
W
i
n
/
W
o
u
t
)
w
e
n
d
=
c
e
i
l
(
(
j
+
1
)
∗
W
i
n
/
W
o
u
t
)
O
u
t
p
u
t
(
i
,
j
)
=
∑
I
n
p
u
t
[
h
s
t
a
r
t
:
h
e
n
d
,
w
s
t
a
r
t
:
w
e
n
d
]
(
h
e
n
d
−
h
s
t
a
r
t
)
∗
(
w
e
n
d
−
w
s
t
a
r
t
)
\begin{aligned} hstart &= floor(i * H_{in} / H_{out})\\ hend &= ceil((i + 1) * H_{in} / H_{out})\\ wstart &= floor(j * W_{in} / W_{out}) \\ wend &= ceil((j + 1) * W_{in} / W_{out}) \\ Output(i ,j) &= \frac{\sum Input[hstart:hend, wstart:wend]}{(hend - hstart) * (wend - wstart)} \end{aligned}
hstarthendwstartwendOutput(i,j)=floor(i∗Hin/Hout)=ceil((i+1)∗Hin/Hout)=floor(j∗Win/Wout)=ceil((j+1)∗Win/Wout)=(hend−hstart)∗(wend−wstart)∑Input[hstart:hend,wstart:wend]
参考自：
AdaptiveAvgPool2D
class CostomAdaptiveAvgPool2D(nn.Layer):
def __init__(self, output_size, input_size):
super(CostomAdaptiveAvgPool2D, self).__init__()
self.output_size = output_size
self.input_size = input_size
def forward(self, x):
H_in, W_in = self.input_size
H_out, W_out = [self.output_size, self.output_size] \
if isinstance(self.output_size, int) \
else self.output_size
out_i = []
for i in range(H_out):
out_j = []
for j in range(W_out):
hs = int(np.floor(i * H_in / H_out))
he = int(np.ceil((i+1) * H_in / H_out))
ws = int(np.floor(j * W_in / W_out))
we = int(np.ceil((j+1) * W_in / W_out))
# print(hs, he, ws, we)
kernel_size = [he-hs, we-ws]
out = F.avg_pool2d(x[:, :, hs:he, ws:we], kernel_size)
out_j.append(out)
out_j = paddle.concat(out_j, -1)
out_i.append(out_j)
out_i = paddle.concat(out_i, -2)
return out_i
用 CostomAdaptiveAvgPool2D 调换 AdaptiveAvgPool2D， 并且指定原图和输出图的大小即可
[14,14] 就是输入图的大小，我这个pp_liteseg 输入shape是
224
×
224
224\times 224
224×224，所以上一层的shape是
[-1, 512, 14, 14]
然后一次过三个AdaptiveAvgPool2D 变成
[-1, 512, 1, 1]
[-1, 512, 2, 2]
[-1, 512, 4, 4]
然后 paddle2onnx 导出即可
paddle2onnx --model_dir . \
--model_filename model.pdmodel \
--params_filename model.pdiparams \
--opset_version 11 \
--save_file output.onnx
最后测试一下：
import paddle
import numpy as np
import paddle.nn as nn
import paddle.nn.functional as F
class CostomAdaptiveAvgPool2D(nn.Layer):
def __init__(self, output_size, input_size):
super(CostomAdaptiveAvgPool2D, self).__init__()
self.output_size = output_size
self.input_size = input_size
def forward(self, x):
H_in, W_in = self.input_size
H_out, W_out = [self.output_size, self.output_size] \
if isinstance(self.output_size, int) \
else self.output_size
out_i = []
for i in range(H_out):
out_j = []
for j in range(W_out):
hs = int(np.floor(i * H_in / H_out))
he = int(np.ceil((i+1) * H_in / H_out))
ws = int(np.floor(j * W_in / W_out))
we = int(np.ceil((j+1) * W_in / W_out))
# print(hs, he, ws, we)
kernel_size = [he-hs, we-ws]
out = F.avg_pool2d(x[:, :, hs:he, ws:we], kernel_size)
out_j.append(out)
out_j = paddle.concat(out_j, -1)
out_i.append(out_j)
out_i = paddle.concat(out_i, -2)
return out_i
# 输入
input_data = np.random.rand(1, 3, 128, 128)
x = paddle.to_tensor(input_data)
adaptive_avg_pool = paddle.nn.AdaptiveAvgPool2D(output_size=1)
pool_out = adaptive_avg_pool(x = x)
costom_avg_pool = CostomAdaptiveAvgPool2D(output_size=(1, 1), input_size=(128, 128))
costom_pool_out = costom_avg_pool(x = x)
>>> costom_pool_out
Tensor(shape=[1, 3, 1, 1], dtype=float64, place=Place(gpu:0), stop_gradient=True,
[[[[0.49809119]],
[[0.49674173]],
[[0.50111742]]]])
>>> pool_out
Tensor(shape=[1, 3, 1, 1], dtype=float64, place=Place(gpu:0), stop_gradient=True,
[[[[0.49809119]],
[[0.49674173]],
[[0.50111742]]]])