深度学习基础知识 使用torchsummary、netron、tensorboardX查看模参数结构

郭庆汝2023-10-10 23:35

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 复制代码 
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 复制代码 
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 复制代码 
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 复制代码 
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 复制代码 
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 复制代码 
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 复制代码 
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 复制代码 
----------------------------------------------------------------
        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 复制代码 
pip install netron onnx

代码实现：

python 复制代码 
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 复制代码 
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 复制代码 
tensorboard --logdir ./logs --port 6006
相关推荐
武子康7 分钟前
大数据-212 数据挖掘 机器学习理论 - 无监督学习算法 KMeans 基本原理 簇内误差平方和
大数据·人工智能·学习·算法·机器学习·数据挖掘
deephub8 分钟前
Tokenformer：基于参数标记化的高效可扩展Transformer架构
人工智能·python·深度学习·架构·transformer
Q81375746014 分钟前
数据挖掘在金融交易中的应用：民锋科技的智能化布局
人工智能·科技·数据挖掘
qzhqbb17 分钟前
语言模型的采样方法
人工智能·语言模型·自然语言处理
qzhqbb20 分钟前
基于 Transformer 的语言模型
人工智能·语言模型·自然语言处理·transformer
___Dream21 分钟前
【CTFN】基于耦合翻译融合网络的多模态情感分析的层次学习
人工智能·深度学习·机器学习·transformer·人机交互
极客代码28 分钟前
【Python TensorFlow】入门到精通
开发语言·人工智能·python·深度学习·tensorflow
义小深31 分钟前
TensorFlow|咖啡豆识别
人工智能·python·tensorflow
Tianyanxiao1 小时前
如何利用探商宝精准营销，抓住行业机遇——以AI技术与大数据推动企业信息精准筛选
大数据·人工智能·科技·数据分析·深度优先·零售
撞南墙者1 小时前
OpenCV自学系列（1）——简介和GUI特征操作
人工智能·opencv·计算机视觉
热门推荐
01【HarmonyOS】HUAWEI DevEco Studio 下载地址汇总02(欧拉)openEuler系统添加网卡文件配置流程、(欧拉)openEuler系统手动配置ipv6地址流程、(欧拉)openEuler系统网络管理说明03组基轨迹建模 GBTM的介绍与实现（Stata 或 R）04【AIGC】重塑未来的科技巨轮05全面解析：构建基于深度学习的安全帽检测系统（UI界面+YOLO代码+数据集）06【经验分享】Ubuntu22.04安装微信(linux官方版)07基于YOLOv10深度学习的CT扫描图像肾结石智能检测系统【python源码+Pyqt5界面+数据集+训练代码】深度学习实战、目标检测08Ubuntu 20.04使用Livox mid 360 测试 FAST_LIO09RAG 实践- Ollama+RagFlow 部署本地知识库10红米手机使用google play