from . import Callback
__all__ = ['History']
[docs]class History(Callback):
"""
Callback that records events into a `History` object.
This callback is automatically applied to
every SuperModule.
"""
def __init__(self, trainer):
super(History, self).__init__()
self.samples_seen = 0.
self.trainer = trainer
def on_train_begin(self, logs=None):
self.epoch_metrics = {
'loss': []
}
self.batch_size = logs['batch_size']
self.has_val_data = logs['has_val_data']
self.has_regularizers = logs['has_regularizers']
if self.has_val_data:
self.epoch_metrics['val_loss'] = []
if self.has_regularizers:
self.epoch_metrics['reg_loss'] = []
def on_epoch_begin(self, epoch, logs=None):
if hasattr(self.trainer._optimizer, '_optimizer'): # accounts for meta-optimizers like YellowFin
self.lrs = [p['lr'] for p in self.trainer._optimizer._optimizer.param_groups]
else:
self.lrs = [p['lr'] for p in self.trainer._optimizer.param_groups]
self.batch_metrics = {
'loss': 0.
}
if self.has_regularizers:
self.batch_metrics['reg_loss'] = 0.
self.samples_seen = 0.
def on_epoch_end(self, epoch, logs=None):
if logs:
self.epoch_metrics['loss'].append(logs['loss'])
if logs.get('val_loss'): # if it exists
self.epoch_metrics['val_loss'].append(logs['val_loss'])
def on_batch_end(self, batch, logs=None):
for k in self.batch_metrics:
self.batch_metrics[k] = (self.samples_seen * self.batch_metrics[k] + logs[k] * self.batch_size) / (self.samples_seen + self.batch_size)
self.samples_seen += self.batch_size
def __getitem__(self, name):
return self.epoch_metrics[name]
def __repr__(self):
return str(self.epoch_metrics)
def __str__(self):
return str(self.epoch_metrics)