aiida-calcmonitor

AiiDA插件，其中包含用于监控正在进行计算的工具。

开发安装

git clone git@github.com:ramirezfranciscof/aiida-calcmonitor.git .
cd aiida-calcmonitor
pip install --upgrade pip
pip install -e .

# These are not available yet!
pip install -e .[pre-commit,testing]  # install extra dependencies
pre-commit install  # install pre-commit hooks
pytest -v  # discover and run all tests

运行监控测试

要运行监控测试，您需要有一个现有的aiida配置文件。您必须在其中运行toymodel代码的计算机上进行设置（需要将aiida_calcmonitor/utils/toymodel_code.sh复制到计算机中，并在AiiDA中设置一个代码）。您可以像这样设置本地主机：

$ verdi computer setup -L localhost -H localhost -T local -S direct -w /scratch/{username}/aiida/ --mpiprocs-per-machine 1 -n
$ verdi computer configure local localhost --safe-interval 5 -n

要运行calcjob监控calcjob，您需要设置一个特殊类型的本地主机，使用以下Mpirun命令：verdi -p <PROFILE_NAME> run（将<PROFILE_NAME>替换为相应的名称）。它还需要一个前置命令，以激活aiida使用的虚拟环境。对于典型的Python虚拟环境，您可以这样做：

source /home/username/.virtualenvs/aiida/bin/activate

一旦所有这些设置完成，您可以使用/examples/example01/submit_everything.py中的示例作为模板，了解如何准备和提交toymodel计算并进行监控。

创建您自己的监控器

要创建您自己的监控器，您需要从MonitorBase类进行子类化。这是一个从Dict派生出来的数据类型，所以您的想法是创建一个子数据类型，该类型具有描述检查过程的方法，然后您将从这个子类型创建一个数据节点，其中包含该过程选项的字典，监控代码将获取该数据节点输入并调用检查方法。

这是一个示例，用于展示可以从父类访问的关键变量（self[...]）以及需要使用的返回值，但结构可以根据需要修改（例如，不需要设置和使用error_detected布尔值，您可以在检查过程中直接返回错误信息）。

class NewMonitorSubclass(MonitorBase):  # pylint: disable=too-many-ancestors
    """Example of monitor for the toy model."""

    def monitor_analysis(self):
        
        sources = self['sources']
        # this contains the mapping to the actual filepaths (see below)

        options = self['options']
        # this contains the specific options for this method (the structure is determined here in this method and the user must know what is expected when constructing it)

        retrieve = self['retrieve']
        # there will also be a list of files to retrieve if the original calculation is killed, but this should probably not be needed here

        internal_naming = sources['internal_naming']['filepath']
        # This is how you access the mapping for the files; now internal_naming contains the actual path to the file it requires

        with open(internal_naming) as fileobj:
            # Here one performs the parsing of the files and checking.
            # Setups the variables error_detected and error_msg or equivalent

        if error_detected:
            return f'An error was detected: {error_msg}'        
            # The calcjob_monitor will interpret string returns as errors and will kill the process
        else:
            return None
            # The calcjob_monitor will interpret None returns to mean everything is going fine

为了可以使用，您还需要将其添加为入口点/插件，例如在pyproject.toml内部。

[project.entry-points."aiida.data"]
"calcmonitor.monitor.toymodel" = "aiida_calcmonitor.data.monitors.monitor_toymodel:MonitorToymodel"
"calcmonitor.monitor.newmonitor" = "aiida_calcmonitor.data.monitors.newmonitor_file:NewMonitorSubclass"

许可

MIT