Updated for the fixed StepLR_Restart learning scheme, now it's consistent with the training strategy published in the PPON paper and additionally, can be trained using MultiStepLR_Restart.
These options are configurable in the JSON file, which now contain both options pre-configured to try to replicate the paper variables.
Updated for the fixed StepLR_Restart learning scheme, now it's consistent with the training strategy published in the PPON paper and additionally, can be trained using MultiStepLR_Restart.
These options are configurable in the JSON file, which now contain both options pre-configured to try to replicate the paper variables.