I switched from Tensorboard to MLFlow. Therefore, still figuring out best practices for utilizing MLFlow. Currently, I am defining the logger in the following way:
logger = MLFlowLogger(experiment_name="MY_EXPERIMENT", save_dir="MY_PATH")
I do not set the run_name
parameter. To the best of my understanding, this should result in MLFlow choosing a random and unique run name. However, MLFlow always produces the same run name capable-auk-759
. Why?
To the best of my knowledge, I have no environment variables set, which could influence the run_name. What could be the root of this problem?
PS:
I also get the following Warning:
WARNING:root:Malformed experiment 'mlruns'. Detailed error Yaml file 'MY_PATH/meta.yaml' does not exist.
Which may or may not have anything to do with the original issue.