onnx模型转换opset版本和固定动态输入尺寸

背景：之前我想把onnx模型从opset12变成opset12，太慌乱就没找着，最近找到了官网上有示例的，大爱onnx官网，分享给有需求没找着的小伙伴们。

1. onnx模型转换opset版本

官网示例：

python 复制代码

import onnx
from onnx import version_converter, helper

# Preprocessing: load the model to be converted.
model_path = "path/to/the/model.onnx"
original_model = onnx.load(model_path)

print(f"The model before conversion:\n{original_model}")

# A full list of supported adapters can be found here:
# https://github.com/onnx/onnx/blob/main/onnx/version_converter.py#L21
# Apply the version conversion on the original model
converted_model = version_converter.convert_version(original_model, <int target_version>)

print(f"The model after conversion:\n{converted_model}")

其github地址如下：

onnx/docs/PythonAPIOverview.md at main · onnx/onnx (github.com)https://github.com/onnx/onnx/blob/main/docs/PythonAPIOverview.md#converting-version-of-an-onnx-model-within-default-domain-aionnx其小伙伴拉到gitee上的地址如下（以防有的小伙伴github打不开）：

docs/PythonAPIOverview.md · meiqicheng/github-onnx-onnx - Gitee.comhttps://gitee.com/meiqicheng1216/onnx/blob/master/docs/PythonAPIOverview.md#converting-version-of-an-onnx-model-within-default-domain-aionnx最后附上完整代码：

python 复制代码

import onnx
from onnx import version_converter, helper

# A full list of supported adapters can be found here:
# https://github.com/onnx/onnx/blob/main/onnx/version_converter.py#L21
# Apply the version conversion on the original model

# Preprocessing: load the model to be converted.
model_path = r"./demo.onnx"
original_model = onnx.load(model_path)
print(f"The model before conversion:\n{original_model}")


converted_model = version_converter.convert_version(original_model, 11)
print(f"The model after conversion:\n{converted_model}")

save_model = model_path[:-5] + "_opset11.onnx"
onnx.save(converted_model, save_model)

2. onnx模型转固定动态输入尺寸

python 复制代码

def change_dynamic_input_shape(model_path, shape_list: list):
    """
    将动态输入的尺寸变成固定尺寸
    Args:
        model_path: onnx model path
        shape_list: [1, 3, ...]
    Returns:

    """
    import os
    import onnx
    model_path = os.path.abspath(model_path)
    output_path = model_path[:-5] + "_fixed.onnx"
    model = onnx.load(model_path)
    # print(onnx.helper.printable_graph(model.graph))
    inputs = model.graph.input  # inputs是一个列表，可以操作多输入~
    # look_input = inputs[0].type.tensor_type.shape.dim
    # print(look_input)
    # print(type(look_input))
    # inputs[0].type.tensor_type.shape.dim[0].dim_value = 1
    for idx, i_e in enumerate(shape_list):
        inputs[0].type.tensor_type.shape.dim[idx].dim_value = i_e
    # print(onnx.helper.printable_graph(model.graph))
    onnx.save(model, output_path)


if __name__ == "__main__":
    model_path = "./demo.onnx"
    shape_list = [1]
    change_dynamic_input_shape(model_path, shape_list)