MLflow UI

Usage

MLflow UI shows the tracking result of the experiments. When we use the log_param or log_metric in ModelClient API, we could view the result in MLflow UI. Below is the example of the usage of MLflow UI.

Example

1. Run the following code in the cluster

from submarine import ModelsClient
import random
import time

if __name__ == "__main__":
  modelClient = ModelsClient()
  with modelClient.start() as run:
      modelClient.log_param("learning_rate", random.random())
      for i in range(100):
        time.sleep(1)
        modelClient.log_metric("mse", random.random() * 100, i)
        modelClient.log_metric("acc", random.random(), i)

2. In the MLflow UI page, you can see the log_param and the log_metric result. You can also compare the training between different workers.