logging to tensorboard Args: epoch: training epoch logs: loss and accuracy lr: learning rate
(self, epoch, logs, lr)
| 22 | super(ContrastTrainer, self).__init__(args) |
| 23 | |
| 24 | def logging(self, epoch, logs, lr): |
| 25 | """ logging to tensorboard |
| 26 | |
| 27 | Args: |
| 28 | epoch: training epoch |
| 29 | logs: loss and accuracy |
| 30 | lr: learning rate |
| 31 | """ |
| 32 | args = self.args |
| 33 | if args.rank == 0: |
| 34 | self.logger.log_value('loss', logs[0], epoch) |
| 35 | self.logger.log_value('acc', logs[1], epoch) |
| 36 | self.logger.log_value('jig_loss', logs[2], epoch) |
| 37 | self.logger.log_value('jig_acc', logs[3], epoch) |
| 38 | self.logger.log_value('learning_rate', lr, epoch) |
| 39 | |
| 40 | def wrap_up(self, model, model_ema, optimizer): |
| 41 | """Wrap up models with apex and DDP |