每次更改验证集(经常)时,我都会创建一个新的“实验”,但每次我都想要基本相同的图表。这是我在训练运行中进行相关比较的最简单方法。有没有办法通过 API 进行设置,这样我就不必每次都点击 UI?
如果您的意思是通过mlflow API“创建实验”,您可以查看以下页面:
https://mlflow.org/docs/latest/rest-api.html#create-experiment
2.0/mlflow/experiments/create
它会返回experiment_id,稍后可以使用。
Python 代码也支持:
import mlflow
mlflow.start_run()
experiment_name = "EXAMPLE"
experiment = mlflow.get_experiment_by_name(experiment_name)
if experiment is None:
experiment_id = mlflow.create_experiment(experiment_name)
else:
experiment_id = experiment.experiment_id
# log metrics
with mlflow.start_run(experiment_id=experiment_id):
mlflow.log_metric("accuracy", 0.1)
mlflow.log_artifact("path/to/artifact/file")
mlflow.end_run()