Why not use tqdm.tqdm?

The tqdm module is a famous library widely used to create dynamic progress bars. LoggerML’s philosophy is to create a similar behavior but oriented in the practice of machine learning. Thus, LoggerML assumes that our python code is organized into epochs and runs through batches of a dataset with values to log. In this context, it gives additional information such as the time remaining until the training is completed.

It also has a much nicer and customizable logging system: the values are organized in tables below the progress bar and use the rich api to style the output.

Finally, the LoggerML logger compute the mean of all numerical values average over epoch (by default) or the overall training. It is particularly interesting as we often need to compute these means for meaningful logging.