当前位置：首页 > news >正文

使用vscode调试transformers源码

news 来源：原创 2025/5/13 9:42:46

简要介绍如何使用vscode调试transformers源码

以源码的方式安装transformers（官方手册为Editable install）

优先参考官方手册

git clone https://github.com/huggingface/transformers.git
cd transformers
pip install -e .

以下展示transformers/examples/pytorch/image-classification样例的调试方法

阅读image-classification的README.md文件，可知需要下载数据集beans和ViT模型（默认为google/vit-base-patch16-224-in21k）,下图是我的数据集目录：
配置launch.json文件，这里面需要根据自己的路径修改一些变量，仅供参考:

{
    // Use IntelliSense to learn about possible attributes.
    // Hover to view descriptions of existing attributes.
    // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
    "version": "0.2.0",
    "configurations": [
    
        {
            "name": "Python Debugger: Current File",
            "type": "debugpy",
            "request": "launch",
            "program": "${file}",
            "console": "integratedTerminal"
        },
        {
            "name": "run_image_classification.py",
            "type": "debugpy",
            "request": "launch",
            "program": "${workspaceFolder}/transformers/examples/pytorch/image-classification/run_image_classification.py",
            "args": [
                "--train_dir", "transformers/examples/pytorch/image-classification/beans/data/train",
                "--validation_dir", "transformers/examples/pytorch/image-classification/beans/data/validation",
                "--output_dir", "./beans_outputs/",
                "--remove_unused_columns","False",
                "--label_column_name", "label",
                "--do_train",
                "--do_eval",
                "--num_train_epochs", "5",
            ],
            "console": "integratedTerminal",
            "justMyCode": false,
            "env": {
              "PYTHONPATH": "${workspaceFolder}"
            }
          },
    ]
}