添加算法
=========

.. contents:: 本节目录
    :depth: 2

命令: ``algorithm``

算法添加分类两类，本地和远程,用 ``--help`` 命令可看到有以下子命令 ``local`` ``remote`` ：

.. code-block:: shell

    anyctl add algorithm --help

.. code-block:: shell

    Usage: anyctl add algorithm [OPTIONS] COMMAND [ARGS]...

        Add local or remote algorithm to local Anylearn project.

    Options:
        --help  Show this message and exit.

    Commands:
        local   Add local algorithm to current project.
        remote  Add remote algorithm by ID to current project.


添加本地算法
---------------

命令: ``local``

+---------------------------+----------+-------------+---------------------------------------------------------+
| 参数及缩写                | 是否必须 | 默认值      | 说明                                                    |
+===========================+==========+=============+=========================================================+
| name                      | True     |             | 算法名称                                                |
+---------------------------+----------+-------------+---------------------------------------------------------+
| -\-dir                    | True     |             | 本地算法文件夹（绝对路径）                              |
+---------------------------+----------+-------------+---------------------------------------------------------+
| -\-entrypoint-training    | True     |             | 启动训练的入口命令                                      |
+---------------------------+----------+-------------+---------------------------------------------------------+
| -\-output-training        | True     |             | 训练输出模型的相对路径（相对于算法目录）                |
+---------------------------+----------+-------------+---------------------------------------------------------+

使用示例：

.. code-block:: shell

    anyctl add algorithm local anyctl_algo --dir D:\anyctl-test\resource\cnn --entrypoint-training "python fashion_mnist.py" --output-training anyctl_algo_result

运行后会有以下输出：

.. code-block:: shell

    [SUCCESS] ADDED  #提示算法添加成功

此时我们用 ``anyctl config ls`` 查看配置项可以看到算法信息已经有了：

.. code-block:: shell

    remote:
        host:
        username:
        password:
    project:
        id:
        name: anyctl_project
        description: A project created by anylearn ctl.
    algorithms:
        anyctl_algo:
            id:
            name: anyctl_algo
            description:
            visibility: 3
            train_params: []
            follows_anylearn_norm: false
            entrypoint_training: python fashion_mnist.py
            output_training: anyctl_algo_result
    datasets: {}
    path:
        algorithm:
            anyctl_algo: D:\anyctl-test\resource\cnn
        dataset: {}

添加远程算法
---------------

我们除了添加本地算法以外还可以添加已经上传到后端的算法，只需要知道后端算法ID即可。

命令: ``remote``

+---------------------------+----------+-------------+---------------------------------------------------------+
| 参数及缩写                | 是否必须 | 默认值      | 说明                                                    |
+===========================+==========+=============+=========================================================+
| id                        | True     |             | 远程算法ID                                              |
+---------------------------+----------+-------------+---------------------------------------------------------+

使用示例：

.. code-block:: shell

    anyctl add algorithm remote ALGOxxx

如果需要配置远程地址和用户信息请参考 :ref:`set_host_user_info` 。

运行后会有以下输出：

.. code-block:: shell

    [SUCCESS] ADDED  #提示算法添加成功

此时我们用 ``anyctl config ls`` 查看配置项可以看到远程算法信息已经有了：

.. code-block:: shell

    ...
    algorithms:
        anyctl_algo:
            id:
            name: anyctl_algo
            description:
            visibility: 3
            train_params: []
            follows_anylearn_norm: false
            entrypoint_training: python fashion_mnist.py
            output_training: anyctl_algo_result
        cli_example_algo:  #远程算法信息
            id: ALGOxxx
            name: cli_example_algo
            description: SDK_QUICKSTART
            visibility: 3
            train_params:
            - name: dataset
              type: dataset
              suggest: 1
            - name: dataset
              type: dataset
              suggest: 1
            - name: model_path
              alias: ''
              description: ''
              type: model
              suggest: 1
            follows_anylearn_norm: false
            entrypoint_training: python fashion_mnist.py
            output_training: model-output
    datasets: {}
    path:
        algorithm:
            anyctl_algo: D:\anyctl-test\resource\cnn
            cli_example_algo:
        dataset: {}