EpistasisLab
Add Callbacks for logging and/or performing custom operations after every generation
Similar to Keras's “callbacks”— we need add a class that can access the training state after every training iteration for better logging and/or performing custom operations after every generation during TPOT optimization. This class is related to a few features in current version (0.9.2) of TPOT (e.g. early stopping and verbosity in general).
Related issue #667
Looking forward to this in TPOT. I think it will clean up the interface quite a bit.
This would be a great feature, e.g., for MLflow to track metrics across generations. What would be required to implement this?
This will be great to have. Would like to integrate with mlflow if possible so that we have more visibility into what has been done, etc.
