The Python TSAI library for time series -- how can I make it less verbose?

Question

I tried to RTFM, here is the documentation:

https://timeseriesai.github.io/tsai/

I found no mention of how to get rid of the table that is output while training. Here is an example of what I do NOT want it to print:

# Assuming you have X, y, and splits
batch_tfms = TSStandardize()
fcst = TSForecaster(X, y, splits=splits, batch_tfms=batch_tfms, bs=nrows, arch=TST, metrics=mae)
fcst.fit_one_cycle(len(splits[0]), 1e-3)

# I do not want this stuff --+
#                            |
#                            v

epoch   train_loss  valid_loss  mae time
0   0.156252    0.108844    0.295480    00:00
1   0.132855    0.075484    0.238252    00:00
2   0.126005    0.058170    0.203647    00:00
3   0.122865    0.048346    0.181954    00:00
4   0.119844    0.043026    0.169333    00:00
5   0.117942    0.039927    0.161866    00:00
6   0.116912    0.037847    0.156890    00:00
7   0.114023    0.036735    0.154284    00:00
8   0.112187    0.036117    0.152822    00:00

Answer 1

The extra output during training is not from the tsai package itself but from the underlying fastai package. Depending on what out you want to show/hide you can use a few different context managers. To hide just the table output you should be able to use the model.no_logging() context manager:

with fcst.no_logging():
    fcst.fit_one_cycle(25, 1e-3)

You can also use the model.no_bar() context manager to disable the progress bar shown.