Visualizing Training

DRDMannTurb relies on TensorBoard for visualizing training results including the behavior of all involved loss terms. Please refer to the TensorBoard API for specifics as well as the LossAggregator arguments for how to set the logging directory and changing metadata.

We also provide a simple utility for plotting out the training progress from a single TensorBoard log.

drdmannturb.common.plot_loss_logs(log_file: str | Path)

Returns a full plot of all loss terms for a specific training log generated by TensorBoard. This is an auxiliary method and should only be used for quick visualization of the training process, the suggested method for visualizing this information is through the TensorBoard API.

Parameters:

log_file (Union[str, Path]) – Full path to training log to be visualized.