安装ONNX包引发Python依赖全冲突，求自动解决方案

你遇到的核心问题是mlagents-envs 0.30.0（硬性要求numpy==1.21.2）与高版本onnx（要求numpy>=1.22）的版本冲突，旧版pip的依赖解析逻辑不完善，导致无法自动协调。以下是几种无需手动试错的自动兼容解决方法：

方法一：升级pip后安装兼容的onnx版本 #

你的pip版本（21.2.3）依赖解析能力有限，先升级到最新版：

python -m pip install --upgrade pip

升级后执行以下命令，pip会自动匹配与现有numpy 1.21.2兼容的onnx版本：

pip install onnx --no-cache-dir

如果仍有问题，可手动指定onnx版本范围（经测试，onnx<=1.12.0均兼容numpy 1.21.2）：

pip install "onnx<=1.12.0"

方法二：用虚拟环境从零构建干净兼容环境 #

虚拟环境可隔离现有依赖干扰，让pip在无残留环境中自动匹配所有兼容版本：

1. 创建虚拟环境：

python -m venv mlagents_venv

2. 激活虚拟环境（Windows系统）：

mlagents_venv\Scripts\activate

3. 一次性安装所有需求包，pip会自动选择相互兼容的版本组合：

pip install torch torchaudio torchvision mlagents protobuf==3.20.3 onnx

方法三：用pip-tools自动生成兼容依赖清单 #

pip-tools可根据你指定的包，自动计算出全兼容的版本组合：

1. 安装pip-tools：

pip install pip-tools

2. 创建requirements.in文件，内容如下：

torch
torchaudio
torchvision
mlagents
protobuf==3.20.3
onnx

3. 编译生成包含所有兼容版本的requirements.txt：

pip-compile requirements.in

4. 根据生成的清单安装所有包：

pip install -r requirements.txt

补充说明 #

你之前尝试pip install onnx==1.11.0无效，大概率是pip缓存了旧依赖信息，加上旧版解析逻辑问题，执行以下命令即可解决：

pip install onnx==1.11.0 --no-cache-dir

错误信息参考 #

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
onnx 1.19.0 requires numpy>=1.22, but you have numpy 1.21.2 which is incompatible.
Successfully installed numpy-1.21.2
WARNING: You are using pip version 21.2.3; however, version 25.2 is available.
You should consider upgrading via the 'C:\Users\me\AppData\Local\Programs\Python\Python310\python.exe -m pip install --upgrade pip' command.

C:\Users\me>pip install --force-reinstall "numpy1.22"
Collecting numpy1.22
Downloading numpy-1.22.0-cp310-cp310-win_amd64.whl (14.7 MB)
|████████████████████████████████| 14.7 MB 6.4 MB/s
Installing collected packages: numpy
Attempting uninstall: numpy
Found existing installation: numpy 1.21.2
Uninstalling numpy-1.21.2:
Successfully uninstalled numpy-1.21.2
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
mlagents-envs 0.30.0 requires numpy==1.21.2, but you have numpy 1.22.0 which is incompatible.

内容的提问来源于stack exchange，提问作者Mich