I'm trying to add metrics to this SummaryWriter, but it's not working. I'm using the add_hparams() function of SummaryWriter, the details of which can be found here: https://pytorch.org/docs/stable/tensorboard.html. I'm doing it this way: writer = SummaryWriter(f'runs/lstm_experiment_final') for e in tqdm(range(num_epochs)): tr_loss, tr_f1, tr_precision, tr_recall = training_loop(model, train_dataloader, loss_function, optimizer, e, writer) val_loss, val_f1, val_precision, val_recall = validation_loop(model, test_dataloader, loss_function, e, writer) metric_dict = {'Loss/train': tr_loss, 'Loss/valid': val_loss, 'F1/train': tr_f1, 'F1/valid': val_f1, 'Precision/train': tr_precision, 'Precision/valid': val_precision, 'Recall/train': tr_recall, 'Recall/valid': val_recall} writer.add_hparams(best_params, metric_dict, global_step=num_epochs-1) writer.close() This is what's happening. enter image description here In other words, hyperparameters are indeed being logged on TensorBoard, but the metric values are not. I hope someone has already seen my problem and knows how to fix this. Thanks in advance. Continue reading...
