MULTIFiT training - log call backs to disk

I followed @pierreguillou ‘s work on Fr and Pt MULTIFiT language models and made good progress creating an Arabic model. One major issue is disappearing notebook cell output (using free resources like colab and Kaggle).
To avoid Buffered data was truncated after reaching the output size limit., is there a way I can capture the output of fit_one_cycle in a file (and specify folder) from this line:
`learn_lm.fit_one_cycle(10, lr, wd=wd, moms=(0.8,0.7), callbacks=[ShowGraph(learn_lm)])’ ?

python - Buffered data was truncated after reaching the output size limit - Stack Overflow

the author says:
Keras framework has a limitation for showing output info in console.
when you see this message, your process is going on in background but you can’t see it.
If you use tensorflow as backend, write a Tensorboard callback in your Keras to see detailed output of your network.

The question is still open how to prevent this output buffer overflows and display the progress and at the same time keep the model running [keras.models.Sequential.fit - with verbose=1 default mode].