MLFlow - 实验跟踪 - 没有这样的文件或目录

Question

. Answers to this question are eligible for a +50 reputation bounty. João Areias wants to to this question.

I want to run experiment tracking on my jupyter notebooks, I've installed mlflow and have set it up as such:

Path("mlruns").mkdir(parents=True, exist_ok=True)
mlflow.set_tracking_uri("file:mlruns")

client = mlflow.MlflowClient()
experiment_id = client.create_experiment(
    EXPERIMENT_NAME,
    tags={
        "author": "me",
    }
)
experiment = client.get_experiment(experiment_id)

Now I want to track it, so I create a plot and run as follows:

with mlflow.start_run(
        experiment_id=experiment_id, 
        run_name="foobar",
        description="Lorem Ipsum dolor sit amet"
    ) as run:

    mlflow.log_figure(fig, "figure_name.png")

but doing so gives me the following error:

FileNotFoundError: [Errno 2] No such file or directory: 
'mlruns\\<experiment_id>\\<run_id>\\tags\\mlflow.source.name'

What am I doing wrong? How can I set it up to track this image?

Answer 1

The Error comes when the file does not exist so your method run and until doesn't generate an image to avoid such this annoying errors use await in python which is new by the way read this blog , I hope my answer helped you and make it to the top of stackoverflow search.