yolov7代码 | model.named_models

前言

了解model.named_models，为剪枝做准备。
剪枝有一些层如果你不想剪掉，那就用需要你会用 model.named_models功能。
先放一段控制剪枝的代码，感受一下

    ignored_layers = [] # 这些层不剪枝
    # ignore output layers 
    # for cfg/training/yolov7-tiny-prune.yaml
    for k, m in model.named_modules():
        if isinstance(m, TSCODE_Detect):
            ignored_layers.append(m.m_cls)
            ignored_layers.append(m.m_reg)
            ignored_layers.append(m.m_conf)
        if isinstance(m, Yolov7_Tiny_E_ELAN_Attention):
            ignored_layers.append(m.att)

1. print(model)

    # Load model
    model = attempt_load(weights, map_location=device)  # load FP32 model
    print(model)

输出网络结构，同学们可以去模型的yaml文件比对一下
yaml文件是模型的结构，打印的model是权重和操作

2. print(model.named_models)

2.1 print(name)

    model = attempt_load(weights, map_location=device)  # load FP32 model
    for name, module in model.named_modules():
        # print(f"{name}:: {module}")
        print("NAME:", name)

2.2 print(module)

    for name, module in model.named_modules():
        print("MODULE:", module)

2.3 print(f"{name}:: {module}")

    # Load model
    model = attempt_load(weights, map_location=device)  # load FP32 model
    for name, module in model.named_modules():
        print(f"{name}:: {module}")

3. hasattr(module, ‘weight’)

    for name, module in model.named_modules():
        if hasattr(module, 'weight'):
            print(module)