APP网址
https://netron.app/
netron是很不的深度学习模型显示工具
netron支持显示大多数的深度学习模型,不支持pytorch生成的pt或者pth文件,但是将这两种文件转为onnx格式,netron是支持的
其他的系统个人还没有尝试过,但是github地址中有,个人暂时只在ubuntu下安装了,很简单
pip install netron
import torch import torch.nn as nn import netron # 定义一个简单的二分类网络 class SimpleNet(nn.Module): def __init__(self): super(SimpleNet, self).__init__() self.conv1 = nn.Sequential( nn.Conv2d(in_channels=3, out_channels=50, kernel_size=3, stride=1, padding=1), nn.ReLU(), nn.MaxPool2d(2) ) self.conv2 = nn.Sequential( nn.Conv2d(in_channels=50, out_channels=200, kernel_size=3, stride=1, padding=1), nn.ReLU(), nn.MaxPool2d(2) ) self.conv3 = nn.Sequential( nn.Conv2d(in_channels=200, out_channels=500, kernel_size=3, stride=1, padding=1), nn.ReLU(), nn.MaxPool2d(2) ) self.conv4 = nn.Sequential( nn.Conv2d(in_channels=500, out_channels=200, kernel_size=3, stride=1, padding=1), nn.ReLU(), nn.MaxPool2d(2) ) self.conv5 = nn.Sequential( nn.Conv2d(in_channels=200, out_channels=50, kernel_size=3, stride=1, padding=1), nn.ReLU(), nn.MaxPool2d(2) ) self.fc = nn.Sequential( nn.Linear(50 * 20 * 11, 5000) ) self.fc1 = nn.Sequential( nn.Linear(5000, 2000) ) self.fc2 = nn.Sequential( nn.Linear(2000, 50) ) self.classifier = nn.Sequential( nn.Linear(50, 2) ) def forward(self, x): x = torch.tensor(x, dtype=torch.float32) x = self.conv1(x) x = self.conv2(x) x = self.conv3(x) x = self.conv4(x) x = self.conv5(x) x = torch.flatten(x, start_dim=1) x = self.fc(x) x = self.fc1(x) x = self.fc2(x) x = self.classifier(x) return x d = torch.rand(1, 3, 640, 360) m = SimpleNet() o = m(d) onnx_path = "dirtyjudgment640320_pattern.onnx" torch.onnx.export(m, d, onnx_path) netron.start(onnx_path)
两张图像中间的conv重复
def convert_model_to_ONNX(input_img_size, input_pth_model, output_ONNX): dummy_input = torch.randn(2, 3, input_img_size[1], input_img_size[0]) model = SimpleNet() #网络结构 state_dict = torch.load(input_pth_model) new_state_dict = OrderedDict() for k, v in state_dict.items(): name = k[7:] # remove `module.` new_state_dict[name] = v model.load_state_dict(new_state_dict) #model.load_state_dict(state_dict) input_names = ["input_image"] #指定输入输出 output_names = ["output_classification"] torch.onnx.export(model, dummy_input, output_ONNX, verbose=True, input_names=input_names, output_names=output_names)
GitHub - lutzroeder/netron: Visualizer for neural network, deep learning, and machine learning models