The number of training epochs seems to be hardcoded to 24 and isn't overwritten anywhere(maybe I'm missing something?), this PR adds a new input argument extra_trainer_config_args to run_cli, I think it's better than directly adding epochs as input because it's more flexible and can be used to specify other trainer args in the future.
The number of training epochs seems to be hardcoded to 24 and isn't overwritten anywhere(maybe I'm missing something?), this PR adds a new input argument
extra_trainer_config_args
torun_cli
, I think it's better than directly addingepochs
as input because it's more flexible and can be used to specify other trainer args in the future.The default behavior is unchanged.