Save the model after each era .
filepath You can include named formatting options , It will fill in the log with the values of the epoch and key （ stay on_epoch_end Transfer in ）.
for example ： If filepath It's weight .{epoch：02d} - {val_loss：.2f} .hdf5, Then the model checkpoint is saved with the era number in the file name and the missing validation .
def __init__(self, filepath: Any, monitor: str = 'val_loss', verbose: int = 0,
save_best_only: bool = False, save_weights_only: bool = False, mode: str =
'auto', period: int = 1) -> None
<> parameter
* filepath：string, The path to save the model file .
* monitor: monitor ： Number to monitor .
* verbose detailed ： Detailed mode ,0 or 1.
* save_best_only： If save_best_only = True, The latest best model based on the number of monitors is not overridden .
* save_weights_only： If True, Only the weight of the model
preservation （model.save_weights（filepath））, Otherwise, save the complete model （model.save（filepath））.
* mode：{auto,min,max} one of . If save_best_only =
True, According to the maximum or minimum of the number of monitors, the current saved file will be overwritten .
about val_acc, It should be max, about val_loss, It should be min etc . In auto mode , Automatically infer direction from monitored quantity name .
* period： Interval between checkpoints （ Number of periods ）.
Technology