Drop tqdm in favor of logger; normalize use of tabulate

[mcioffi]

The Updates ✏️

TL;DR — Future web app (PR #25) use websockets for routing training data to the front-end, and tqdm adds complexity to that with stdout behavior over sockets, so this proposes to remove it; also included are updates to tabulate printing and a decorate function

• Trainer: Remove tqdm dependency
• Trainer: Reverts to logger.info for multi-processing tasks like train (multiple), gridsearch, and featuresearch
• Trainer: Normalize use of tabulate across all tasks — train (single/multiple) now gets this treatment (examples below)
• Trainer: New decorator fn called set_redirect_log_stream, to be used in web app

$ python3 train.py train --database train/data/training.sqlite3

[INFO] (training_utils) Loading and transforming training data.
[INFO] (training_utils) 81,272 usable vectors.
[INFO] (training_utils) 45 discarded due to OTHER labels.
[INFO] (train_model) 905609147 is the random seed used for the train/test split.
[INFO] (train_model) 65,017 training vectors.
[INFO] (train_model) 16,255 testing vectors.
[INFO] (train_model) Training model with training data.
[INFO] (train_model) Evaluating model with test data.

+------------------------+---------------------------+
| Sentence-level results | Word-level results        |
+------------------------+---------------------------+
| Accuracy: 94.42%       | Accuracy: 97.84%          |
|                        | Precision (micro) 97.82%  |
|                        | Recall (micro) 97.84%     |
|                        | F1 score (micro) 97.83%   |
+------------------------+---------------------------+

$ python3.12 train.py multiple --database train/data/training.sqlite3 --runs 2

[INFO] (training_utils) Loading and transforming training data.
[INFO] (training_utils) 81,272 usable vectors.
[INFO] (training_utils) 45 discarded due to OTHER labels.
[INFO] (train_model) Queued for 2 separate runs. This may take some time.
[INFO] (train_model) 1st run completed
[INFO] (train_model) 2nd run completed

+---------+---------------------+-------------------+-----------+
| Run     | Word/Token accuracy | Sentence accuracy | Seed      |
+---------+---------------------+-------------------+-----------+
| 1st     | 97.75%              | 94.36%            | 94545036  |
| 2nd     | 98.01%              | 94.83%            | 125339459 |
| -       | -                   | -                 | -         |
| Average | 97.88% ± 0.54%      | 94.60% ± 0.98%    | 125339459 |
| -       | -                   | -                 | -         |
| Best    | 98.01%              | 94.83%            | 125339459 |
| Worst   | 97.75%              | 94.36%            | 94545036  |
+---------+---------------------+-------------------+-----------+

[strangetom]

Thanks @mcioffi

A few minor issues:

1. Can you run the pre-commit hooks to fix the formatting issues
2. Using the fancy_grid style for the tables makes for better readability when the table columns are wide (e.g. when grid search results)

[strangetom]

Perfect, thanks!