Logging and Visualization

Now, the training process has been completed. The log information will be stored in the save_dir directory. Assuming that:

  • The filename of the training configuration file is: baseline.toml

  • The value of the save_dir parameter in the baseline.toml is sdnn_delays/exp

Then, the log information will be stored in the sdnn_delays/exp/baseline directory, which contains the following information:

.
├── baseline.log
├── checkpoints
├── config__2023_01_13--10_27_42.toml
├── enhanced
└── tb_log
    └── events.out.tfevents.1673576862.VM-97-67-ubuntu.3747605.0

  • baseline.log: the log information.

  • checkpoints/: model checkpoints.

  • config__2023_04_13--10_27_42.toml: a backup of the training configuration file.

  • enhanced: the enhanced audio files when running in test mode

  • tb_log/: tensorBoard log information, we can visualize it through TensorBoard

Currently, we only support TensorBoard for visualization. Assuming that the value of the save_dir parameter in the basline.toml is sdnn_delays/exp, then we can use the following command to visualize the log information:

tensorboard --logdir sdnn_delays/exp --bind_all