How to save a trained model to remote mlflow server?

Question

I have a server that mlflow is running on port 8791 with --default-artifact-root set to file:/home/data/mlflow.

This server does not have a GPU, so I trained the model on another server and now I want to save it using mlflow.

I did this:

mlflow.set_tracking_uri('http://MyServerIP:8791')
mlflow.set_experiment('experiment1')
with mlflow.start_run() as run:
    mlflow.log_metric('val loss', 0.23656)
    mlflow.log_metric('val accuracy', 0.221)
    mlflow.log_metric('val precision', 0.545)
    mlflow.log_metric('val recall', 0.221)
    mlflow.log_metric('val f1', 0.4554)
    mlflow.pytorch.log_model(model, "model")
    model_uri: str = f"runs:/{run.info.run_id}/model"
    registered_model_version: ModelVersion = mlflow.register_model(model_uri, mlflow_model_name)

The model appears to be saved as I can see the metrics, experiment, and model name in the MLflow UI. Also, in the model version table, there is a row with its source value as file:///home/data/mlflow/4/19b1f0fa0c35492096568e38c095835e/artifacts/model.

However, this path does not exist on my server at all. What could be the problem? How can I save the model into mlflow from the external server?