Add Docker support for running onnx-modifier

[helloworld77]

Description

Summary

This PR adds the ability to run onnx-modifier in a Docker container, so users can start the service without installing Python or dependencies locally. The documentation in README is updated with a new section "launch from Docker container".

Motivation

• Users in containerized or restricted environments can run onnx-modifier without a local Python setup.
• Deployment and sharing become easier (e.g. on servers or CI).
• Keeps the same workflow as other launch methods (clone → run → open in browser).

Changes

1. Dockerfile (new)

   • Base image: Python 3.9 slim.
   • Install dependencies from requirements.txt.
   • Default command: python entry.py --host 0.0.0.0 --port 5000 so the app is reachable from outside the container.

2. README.md

   • New section "launch from Docker container" with:
     • docker build and docker run commands.
     • Note that Flask will show http://0.0.0.0:5000/ and http://127.0.0.1:5000/; opening either in the browser launches onnx-modifier.

3. onnx_modifier/flask_server.py

   • entry.py (or the script that starts the server) is assumed to support --host and --port so the server can bind to 0.0.0.0:5000 inside the container. Any small changes in this file are to support that and/or fix download behavior when running in Docker (e.g. using absolute path for send_file if needed). Models can now download directly through the browser as a standard file.

How to use (for reviewers/users)

git clone https://github.com/ZhangGe6/onnx-modifier.git
cd onnx-modifier
docker build -t onnxmodifier .
docker run -d --restart always --pids-limit -1 --security-opt seccomp=unconfined -p 5000:5000 --name onnxmodifier onnxmodifier

Then open http://127.0.0.1:5000/ (or http://0.0.0.0:5000/) in a browser.

Note for maintainers

• If you prefer to use the official Docker Hub Python image in the upstream repo, the Dockerfile base image can be set to python:3.9-slim and pip install can be run without a custom index (e.g. remove -i https://pypi.tuna.tsinghua.edu.cn/simple for a generic Dockerfile).

---

修改说明

主要特征：支持通过 Docker 容器启动 onnx-modifier

简述

本 PR 增加通过 Docker 运行 onnx-modifier 的方式，并在 README 中新增「launch from Docker container」小节，包含构建与运行命令。使用方式与从源码启动一致：构建镜像 → 运行容器 → 在浏览器中打开输出的地址即可使用。

主要修改

1. 新增 Dockerfile：基于 Python 3.9 slim，安装 requirements.txt，默认以 --host 0.0.0.0 --port 5000 启动，便于从宿主机访问。
2. 更新 README.md：增加 Docker 构建与运行示例，并说明浏览器可访问 http://0.0.0.0:5000/ 或 http://127.0.0.1:5000/。
3. onnx_modifier/flask_server.py：确保支持 --host/--port 并在容器内正确绑定；如有需要，修正下载逻辑（如使用绝对路径）以在 Docker 环境下正常工作。修改static/index.js 和 flask_server.py，支持浏览器直接下载修改后的模型作为标准文件（同时保留原有的存储到指定目录）。

使用步骤

见上方英文版中的 docker build 与 docker run 命令，启动后在浏览器打开 http://127.0.0.1:5000/ 即可。

[ZhangGe6]

@helloworld77 Thanks for the contribution. What are the necessities of the modifications in static/index.js and flask_server.py?

[helloworld77]

@helloworld77 Thanks for the contribution. What are the necessities of the modifications in static/index.js and flask_server.py?

"Sorry for the late reply. The modifications in static/index.js and flask_server.py enable a direct browser-side download feature.

Currently, the tool saves modified models into a predefined folder on the server. With these changes, users can now download the modified model directly through their browser as a standard file download, which is more flexible for remote deployments."