深度学习基础知识 使用torchsummary、netron、tensorboardX查看模参数结构
1、直接打印网络参数结构
python
import torch.nn as nn
from torchsummary import summary
import torch
class Alexnet(nn.Module):
def __init__(self):
super().__init__()
self.net = nn.Sequential(nn.Conv2d(3, 96, kernel_size=11, stride=4, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(96, 256, kernel_size=5, padding=2), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(256, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 256, kernel_size=3, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Flatten(), nn.Linear(256 * 5 * 5, 4096), nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(4096, 4096), nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(4096, 10))
def forward(self, X):
return self.net(X)
if __name__=="__main__":
device=torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model=Alexnet().to(device)
print(model)
# summary(model,(3,224,224),16)结果输出：
python
Alexnet(
(net): Sequential(
(0): Conv2d(3, 96, kernel_size=(11, 11), stride=(4, 4), padding=(1, 1))
(1): ReLU()
(2): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)
(3): Conv2d(96, 256, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
(4): ReLU()
(5): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)
(6): Conv2d(256, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(7): ReLU()
(8): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(9): ReLU()
(10): Conv2d(384, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(11): ReLU()
(12): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)
(13): Flatten(start_dim=1, end_dim=-1)
(14): Linear(in_features=6400, out_features=4096, bias=True)
(15): ReLU()
(16): Dropout(p=0.5, inplace=False)
(17): Linear(in_features=4096, out_features=4096, bias=True)
(18): ReLU()
(19): Dropout(p=0.5, inplace=False)
(20): Linear(in_features=4096, out_features=10, bias=True)
)
)上述方案存在的问题是：当网络参数设置存在错误时，无法检测出来
2、采用torchsummary检测、查看模型参数结构
安装torchsummary
python
pip install torchsummary通常采用torchsummary打印网络结构参数时，会出现以下问题
代码：
python
import torch.nn as nn
from torchsummary import summary
class Alexnet(nn.Module):
def __init__(self):
super().__init__()
self.net = nn.Sequential(nn.Conv2d(3, 96, kernel_size=11, stride=4, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(96, 256, kernel_size=5, padding=2), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(256, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 256, kernel_size=3, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Flatten(), nn.Linear(256 * 5 * 5, 4096), nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(4096, 4096), nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(4096, 10))
def forward(self, X):
return self.net(X)
net = Alexnet()
print(summary(net, (3, 224, 224), 8))报错内容如下：
python
RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same报错原因分析：
在使用torchsummary可视化模型时候报错，报这个错误是因为类型不匹配，根据报错内容可以看出Input type为torch.FloatTensor（CPU数据类型），而weight type（即网络权重参数这些）为torch.cuda.FloatTensor（GPU数据类型）
解决方案：
将model传到GPU上便可。将代码如下修改便可正常运行
python
if __name__ == "__main__":
from torchsummary import summary
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model = UNet().to(device) # modify
print(model)
summary(model, input_size=(3, 224, 224))整体代码：
python
import torch.nn as nn
from torchsummary import summary
import torch
class Alexnet(nn.Module):
def __init__(self):
super().__init__()
self.net = nn.Sequential(nn.Conv2d(3, 96, kernel_size=11, stride=4, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(96, 256, kernel_size=5, padding=2), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(256, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 256, kernel_size=3, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Flatten(), nn.Linear(256 * 5 * 5, 4096), nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(4096, 4096), nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(4096, 10))
def forward(self, X):
return self.net(X)
if __name__=="__main__":
device=torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model=Alexnet().to(device)
# print(model)
summary(model,(3,224,224),16) # 16:表示传入的数据批次打印结果：
python
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
Conv2d-1 [16, 96, 54, 54] 34,944
ReLU-2 [16, 96, 54, 54] 0
MaxPool2d-3 [16, 96, 26, 26] 0
Conv2d-4 [16, 256, 26, 26] 614,656
ReLU-5 [16, 256, 26, 26] 0
MaxPool2d-6 [16, 256, 12, 12] 0
Conv2d-7 [16, 384, 12, 12] 885,120
ReLU-8 [16, 384, 12, 12] 0
Conv2d-9 [16, 384, 12, 12] 1,327,488
ReLU-10 [16, 384, 12, 12] 0
Conv2d-11 [16, 256, 12, 12] 884,992
ReLU-12 [16, 256, 12, 12] 0
MaxPool2d-13 [16, 256, 5, 5] 0
Flatten-14 [16, 6400] 0
Linear-15 [16, 4096] 26,218,496
ReLU-16 [16, 4096] 0
Dropout-17 [16, 4096] 0
Linear-18 [16, 4096] 16,781,312
ReLU-19 [16, 4096] 0
Dropout-20 [16, 4096] 0
Linear-21 [16, 10] 40,970
================================================================
Total params: 46,787,978
Trainable params: 46,787,978
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 9.19
Forward/backward pass size (MB): 163.58
Params size (MB): 178.48
Estimated Total Size (MB): 351.25
----------------------------------------------------------------3、采用netron检测、查看模型参数结构
安装netron与onnx
python
pip install netron onnx代码实现：
python
import torch.nn as nn
import netron
import torch
from onnx import shape_inference
import onnx
class Alexnet(nn.Module):
def __init__(self):
super().__init__()
self.net = nn.Sequential(nn.Conv2d(3, 96, kernel_size=11, stride=4, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(96, 256, kernel_size=5, padding=2), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(256, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 256, kernel_size=3, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Flatten(), nn.Linear(256 * 5 * 5, 4096), nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(4096, 4096), nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(4096, 10))
def forward(self, X):
return self.net(X)
if __name__=="__main__":
device=torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model=Alexnet()
temp_image=torch.rand((1,3,224,224))
# 1、利用torch.onnx.export，先将模型导出为onnx格式的文件，保存到本地./model.onnx
torch.onnx.export(model=model,args=temp_image,f='model.onnx',input_names=['image'],output_names=['feature_map'])
# 2、加载进onxx模型，并推理,然后再保存覆盖原先模型
onnx.save(onnx.shape_inference.infer_shapes(onnx.load("model.onnx")),"model.onnx")
netron.start('model.onnx')运行后，显示结构：
3、使用tensorboardX
代码实现：
python
import torch
import torch.nn as nn
from tensorboardX import SummaryWriter as SummaryWriter
class Alexnet(nn.Module):
def __init__(self):
super().__init__()
self.net = nn.Sequential(nn.Conv2d(3, 96, kernel_size=11, stride=4, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(96, 256, kernel_size=5, padding=2), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(256, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 256, kernel_size=3, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Flatten(), nn.Linear(256 * 5 * 5, 4096), nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(4096, 4096), nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(4096, 10))
def forward(self, X):
return self.net(X)
net = Alexnet()
img = torch.rand((1, 3, 224, 224))
with SummaryWriter(log_dir='logs') as w:
w.add_graph(net, img)运行后，会在本地生成一个log日志文件
在命令行运行以下指令：
python
tensorboard --logdir ./logs --port 6006