After the model is trained, I use tensorboard to find that the reward mean goes up and down. How do I automatically identify the checkpoint (number) that gives the peak in the reward mean/max/min?
With this peak checkpoint, I can further train other models starting from this checkpoint.
Hi,
After the model is trained, I use tensorboard to find that the reward mean goes up and down. How do I automatically identify the checkpoint (number) that gives the peak in the reward mean/max/min?
With this peak checkpoint, I can further train other models starting from this checkpoint.
Best, Yulin