loggerml 0.2.9

Log your ml training in the console in an attractive way.

LoggerML - Machine Learning Logger in the console

Log your Machine Learning training in the console in a beautiful way using rich✨ with useful information but with minimal code.

Installation

In a new virtual environment, install simply the package via pipy:

pip install loggerml

Supported platforms

This package assume that you are using a terminal that support ANSI escape sequences. See here for supported platforms. All Unix and Emacs distribution are supported as well as Windows but only on some machine (Windows 11 seems to work but not Windows 10).

The quick test to know if your terminal support ANSI escape sequence is to run the following command in your terminal:

python -c "print('\x1B')"

It should print an empty line.

Quick start

Minimal usage

Integrate the LogML logger in your training loops! For instance for 4 epochs, 20 batches per epoch and a log interval of 2 batches:

from logml import Logger

logger = Logger(
    n_epochs=4,
    n_batches=20,
    log_interval=2,
)
for _ in range(4):
    logger.start_epoch()  # Indicate the start of a new epoch
    for _ in range(20):
        logger.start_batch()  # Indicate the start of a new batch
        logger.log({'loss': 0.54321256, 'accuracy': 0.85244777})

Yields:

Epoch 1/4, batch 20/20
[================================================][100%]
[global 00:00:02 > 00:00:06 | epoch 00:00:02 > 00:00:00]
loss: 0.5432 | accuracy: 0.8524 |

Epoch 2/4, batch 8/20
[=================>                              ][40%]
[global 00:00:03 > 00:00:05 | epoch 00:00:01 > 00:00:01]
loss: 0.5432 | accuracy: 0.8524 |

Advanced usage

Now you can customize the logger with your own styles and colors. You can set the default configuration at the initialization of the logger and then you can override it during log. You can also log the averaged value over the epoch. For instance:

logger = Logger(
    n_epochs=4,
    n_batches=20,
    styles='yellow',
    digits={'accuracy': 2},
    average=['loss'],  # loss will be averaged over the current epoch
    bold_keys=True,
    show_time=False,  # Remove the time bar
)
for _ in range(4):
    logger.start_epoch()
    for _ in range(20):
        logger.start_batch()
        # Overwrite the default style for "loss" and add a message
        logger.log(
            {'loss': 0.54321256, 'accuracy': 85.244777},
            styles={'loss': 'italic red'},
            message="Training is going well?\nYes!",
        )

Yields:

Epoch 1/4, batch 20/20
[================================================][100%]
loss: 0.5432 | accuracy: 85 |

Epoch 2/4, batch 7/20
[=================>                              ][35%]
[global 00:00:03 > 00:00:05 | epoch 00:00:01 > 00:00:01]
loss: 0.5432 | accuracy: 85 |
Training is going well?
Yes!

With "loss: 0.5432" in italic red, "accuracy: 85" in yellow and both keys in bold.

Don't know the number of batches in advance?

If you don't have the number of batches in advance, you can initialize the logger with n_batches=None. The progress bar is replaced by a cyclic animation. The eta times are not know at the first epoch but was estimated after the second epoch.

How to contribute

For development, install the package dynamically and dev requirements with:

pip install -e .
pip install -r requirements-dev.txt

Everyone can contribute to LogML, and we value everyone’s contributions. Please see our contributing guidelines for more information 🤗

License

Copyright (C) 2023 Valentin Goldité

This program is free software: you can redistribute it and/or modify it under the terms of the MIT License. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

This project is free to use for COMMERCIAL USE, MODIFICATION, DISTRIBUTION and PRIVATE USE as long as the original license is include as well as this copy right notice at the top of the modified files.