onnx.onnx_cpp2py_export.checker.ValidationError: Nodes in a graph must be topologically sorted的可能原因

onnx.onnx_cpp2py_export.checker.ValidationError: Nodes in a graph must be topologically sorted, however input 'model1_onnx::Conv_286' of node:
name: Conv_0 OpType: Conv
 is not output of any previous nodes.

上述报错的可能原因是，只修改了onnx中每个node的input，output属性和output的属性。而没有修改weight和bias ，running mean等参数，正确修改方法:

def rename_model(model, prefix='aaa'):
    for node in model.graph.node:
        node.name=f"{prefix}_{node.name}"

        for index in range(len(node.input)):
            node.input[index] = f"{prefix}_{node.input[index]}"
        for index in range(len(node.output)):
            node.output[index] = f"{prefix}_{node.output[index]}"


    for index in range(len(model.graph.initializer)):
        model.graph.initializer[index].name=f"{prefix}_{model.graph.initializer[index].name}"

    # rename input
    model.graph.input[0].name = f"vuy_img"
    model.graph.node[0].input[0] = f"vuy_img"

    # rename output
    for index in range(len(model.graph.output)):
        model.graph.output[index].name = f"{prefix}_{model.graph.output[index].name}"

可以通过调用下面的函数，看onnx格式是否正确：（如果不能被化简，则说明onnx格式不正确）

def simplify_onnx(onnx_path):
    """

    Args:
        onnx_path ():

    Returns:

    """
    onnx_model = onnx.load(onnx_path)  # load onnx model
    model_simp, check = simplify(onnx_model)
    assert check, "Simplified ONNX model could not be validated"

    save_path = onnx_path
    onnx.save(model_simp, save_path)
    print(f"Simplified onnx save path {save_path}")
    return save_path