相同的MLflow项目在CLI中运行，但API失败

问题描述

我有一个简单的.py项目来测试MLProject功能。

使用CLI：

mlflow run -e clean_tables -P table=COSKAN -P operation=TEST  \
   git@<my-git-host>:data-scientist/utilities.git

运行完美

但是如果使用API：

params = {"table": "COSKAN","operation": "TEST"}
mlflow.projects.run(uri="git@<my-git-host>:data-scientist/utilities.git",entry_point="clean_tables",parameters=params,experiment_name="clean_table",synchronous=True)

它会引发运行时错误。

这部分是Jupyter Notebook信息。

2020/09/10 10:37:25 INFO mlflow.projects.utils: === Fetching project from git@<my-git-host>:data-scientist/utilities.git into /tmp/tmpmfpx05mp ===
2020/09/10 10:37:27 INFO mlflow.projects: === Created directory /tmp/tmpros5uq7p for downloading remote URIs passed to arguments of type 'path' ===
2020/09/10 10:37:27 INFO mlflow.projects: === Running command 'source /z/miniconda/bin/../etc/profile.d/conda.sh && conda activate mlflow-cce1736b779f94bfae0b1056bd02fa8b612d9412 1>&2 && python clean_tables.py --table COSKAN --operation TEST' in run with ID '11a65566b0714279a45dfbaa17689d6e' === 
---------------------------------------------------------------------------
This part is the stack trace:
ExecutionException                        Traceback (most recent call last)
<ipython-input-78-d8443b2b2576> in <module>
----> 1 mlflow.projects.run(uri="git@<my-git-host>:data-scientist/utilities.git",synchronous=True)

<my_home>/environment/lib/python3.6/site-packages/mlflow/projects/__init__.py in run(uri,entry_point,version,parameters,docker_args,experiment_name,experiment_id,backend,backend_config,use_conda,storage_dir,synchronous,run_id)
    281         synchronous=synchronous)
    282     if synchronous:
--> 283         _wait_for(submitted_run_obj)
    284     return submitted_run_obj
    285 

<my_home>/environment/lib/python3.6/site-packages/mlflow/projects/__init__.py in _wait_for(submitted_run_obj)
    298         else:
    299             _maybe_set_run_terminated(active_run,"Failed")
--> 300             raise ExecutionException("Run (ID '%s') Failed" % run_id)
    301     except KeyboardInterrupt:
    302         _logger.error("=== Run (ID '%s') interrupted,cancelling run ===",run_id)

ExecutionException: Run (ID '11a65566b0714279a45dfbaa17689d6e') Failed

解决方法

暂无找到可以解决该程序问题的有效方法，小编努力寻找整理中！

如果你已经找到好的解决方法，欢迎将解决方案带上本链接一起发送给小编。

小编邮箱:dio#foxmail.com (将#修改为@）

mlflow