Google Code-in

tf.keras makes it extremely easy to get started with deep learning. It also provides an easy interface for researchers who are interested in having more fine-grained control over their training loops.

Logging is very important for keeping track of the progress a model is making. A good logger, in this case, should include information about the following:

• Number of samples processed in the current batch
• Current epoch
• Time taken to execute the current epoch
• Any metric supplied to the model in its compilation stage

Currently, tf.keras does provide information about all the above-mentioned points, but can we go one step further? Can we facilitate the use of progress bars in the callbacks to make logs even more interactive and interesting?

The purpose of this task is to design a callback that could make the use of animated progress bar-based logging possible.

Please refer to this example. The task is to develop something similar to this. Note that the animated graph is not mandatory.