NEWSLETTER
Learn how to set up an efficient MLflow environment to track your experiments, compare, and choose the best model for deployment.
Training and tuning multiple models is a basic task for every computer vision researcher. Even for the easiest ones, we perform a hyperparameter search to find the optimal way to train the model on our custom dataset. Data augmentation techniques (which already include many different options), the choice of optimizer, the learning rate and the model itself. Is it the best architecture for my case? Should I add more layers, change the architecture and there will be many more questions waiting to be asked and searched?
While searching for an answer to all these questions, I used to save the log files of the model training process and output checkpoints in different folders on my local, rename the output directory every time I ran a training and compare the final metrics manually once. for one. Approaching the experiment tracking process in such a manual way has many disadvantages: it is old school, time- and energy-consuming, and error-prone.
In this blog post, I'll show you how to use MLflow, one of the best tools to keep track of your experiment, allowing you to record any information you need, visualize and compare the different training experiments you've done, and decide which training is best. . Optimal choice in a user (and eye) friendly environment!
