MLFLow
MLFlow is a tool for tracking and analyzing machine learning experiments. It supports all major machine learning frameworks and is open-source and free. MLFlow helps with four key pain points in the ML life cycle: experiment tracking, reproducibility, packaging, and registry.
MLSync can help bring all the functionality of MLFlow to your productivity tools to better help plan, manage, and execute on your machine learning projects.
MLFlow Configuration
MLSync requires the following configurations of the MLFlow environment:
MLFlow tracking URI: The URI of the MLFlow tracking server. The default value is
http://localhost:5000
. You can set the uri in your code withmlflow.set_tracking_uri(uri)
. You can pass to MLSync in two ways:--mlflow-uri
: The uri is passed as an argument tomlsync
.In your
config.yaml
file, under themlflow
section, set theuri
key to the uri.
If you pass the uri in the command line, it will override the uri in the config file and store the uri in the config file for future runs.
Below is an example config.yaml
file for MLFlow:
mlflow: uri: http://localhost:5000