logging to tensorboard Args: epoch: training epoch logs: loss and accuracy lr: learning rate train: True of False
(self, epoch, logs, lr=None, train=True)
| 17 | super(LinearTrainer, self).__init__(args) |
| 18 | |
| 19 | def logging(self, epoch, logs, lr=None, train=True): |
| 20 | """ logging to tensorboard |
| 21 | |
| 22 | Args: |
| 23 | epoch: training epoch |
| 24 | logs: loss and accuracy |
| 25 | lr: learning rate |
| 26 | train: True of False |
| 27 | """ |
| 28 | args = self.args |
| 29 | if args.rank == 0: |
| 30 | pre = 'train_' if train else 'test_' |
| 31 | self.logger.log_value(pre+'acc', logs[0], epoch) |
| 32 | self.logger.log_value(pre+'acc5', logs[1], epoch) |
| 33 | self.logger.log_value(pre+'loss', logs[2], epoch) |
| 34 | if train and (lr is not None): |
| 35 | self.logger.log_value('learning_rate', lr, epoch) |
| 36 | |
| 37 | def wrap_up(self, model, classifier): |
| 38 | """Wrap up models with DDP |