Hello,
I really enjoy this! I hope that adding the following features would be more helpful.
1: The name of the training data directory other than /image can be selected by config.yml. Otherwise, I kept changing the folder name to "image".
2: plotting training/validation losses
3: The model folder is full of saved checkpoints for each epoch (already 400 GB in my case, too painful with someone using a compact PC setting). config.yml to specify how to keep the checkpoint data (save all, save the best one only, save the last one only, etc.).
Thanks!!
Hello,
I really enjoy this! I hope that adding the following features would be more helpful.
1: The name of the training data directory other than /image can be selected by config.yml. Otherwise, I kept changing the folder name to "image".
2: plotting training/validation losses
3: The model folder is full of saved checkpoints for each epoch (already 400 GB in my case, too painful with someone using a compact PC setting). config.yml to specify how to keep the checkpoint data (save all, save the best one only, save the last one only, etc.).
Thanks!!