A simple logger for notebooks or repl

[ttumiel]

🚀 Feature

A basic logger that can display results in a table in the repl or in a notebook as the network trains and saves the results to a csv file.

Motivation

A lightweight logger is useful for environments where you might not have port access or for interactive experimentation.

Pitch

I created a logger at this gist that does this: https://gist.github.com/ttumiel/32e6a73d206f4df55aa73d6d4ecdf0c4

Alternatives

In #1803, a CSV logger is proposed which would perform similarly except it would not display interactively. If the csv logger is implemented, a subclass could display the csv in the repl.
Furthermore, lightning doesn't depend on pandas which my implementation above uses. I could remove pandas, though, and build the csv manually.

[github-actions]

Hi! thanks for your contribution!, great first issue!

[ttumiel]

A nice and clean alternative I found is using a callback such as the PrintTableMetricsCallback from bolt:

from pl_bolts.callbacks import PrintTableMetricsCallback
import pytorch_lightning as pl

trainer = pl.Trainer(callbacks=[PrintTableMetricsCallback()])

# loss│train_loss│val_loss│epoch
# ──────────────────────────────
# 2.2541470527648926│2.2541470527648926│2.2158432006835938│0

Since this issue is about displaying the logs in a table in the repl I'll close this. #1803 is still open though for a csv logger, which potentially could be implemented in a similar way.