Overview

MLOPs (Machine Learning OPerations) is a set of practises to reliably deploy and maintain machine learning models in production environments. Training deep learning models can be a very tedious process that involves running multiple iterations of training with different combinations of hyperparameters. This may result in several experiments and metrics to track. To help streamline experiment and metric tracking, TAO integrates with the following third-party MLOPS services:

• Weights and Biases (wandb)

• ClearML

• Tensorboard through FTMS

Note

To integrate TAO with a third-party MLOPs service, you will need to set up an account with the respective service. Follow the Log in instructions on the page for the respective service. If you do not wish to set up an account, follow the instuctions for Tensorboard through FTMS.

Warning

Protection of intellectual properties on external MLOps is the responsability of the account owner. Please follow MLOPs security recommendations.