Segmind Track is a client library written in python that helps you keep track of your deep learning experiments. Track library can log your runs, output metrics, hyperparameters, and artifacts.
Key scalar metrics such as loss, precision, recall, IoU, etc. help you understand and monitor your training run. You may also want to compare these metrics across multiple training runs to help debug and improve your model.
Segmind also logs system metrics including CPU, Memory, and GPU usage to help you monitor your hardware usage.
Log your hyperparameters and other configurations that do not change during the run. This helps in analysing the effect of each parameter, correlation between parameters, and reproducing your experiments in the future.
Log important files such as checkpoint files, predictions, tables, Tensorboard files, etc. to store them for analysing your training runs and retrieving them in the future.
Check out Segmind on Github.