logml

• ✨ Allow multiple logs on same step

🆙 Major release 1.0 !

LoggerML is ready to be used in your ML projects!

Project features:

• 🎉 Working logger class logml.Logger that log any bool, str, int, float values with rich style, using rich.Live.
• ✨ Possibility to have multiple logger (for instance one for training and one for validation).
• ✨ Compute the mean of the values (and logged them if needed) over epoch or over the overall training.
• ✨ Default configuration for style, size and whether to average. The configuration can be override when logging.
• ✨ Allow regex matching for log configuration.
• ✨ Easy to use with experiment tracker like Weights and Biases (with Logger.get_vals)
• 📝 Complete documentation with Read the Docs.
• 👷 Automatic pypi build on new tags.
• 🎨 10/10 pylint score + clean ruff (all flake8 + extensions) and mypy.
• ✔️ Integration test and 100% coverage.

Since last version:

• 🚀 Improve behavior when Logger.log is used with Logger.new_epoch/batch
• 📝 Add precision on READMEs and documentation pages

• ✨ Add regex matching for config log
• 📝 Add features documentation

• ✨ Add logger.tqdm feature to simplify code
• 📝 Add sections id doc (vs_tqdm and with_tracker)

📝 ✨ Add a nice documentation available at https://logml.readthedocs.io/en/latest/

• 🚀 Use rich.live and renderable groups to improve libaility
• ✔️ Test with bool and str values in log
• 📝 Update readmes with new features

🔖 Add automatic __version__ with setuptools_scm

• ✨ Add color functionalities: name, name_style and highlight (with rich)
• ⬇️ Allow more flexible versions
• 📝 Add Support plateforms section in readmes

• 🚑 Fix pypi build
• ✨ Now manage small terminal width or long values to log

✨ Functional Logger class that log the epoch and batch number, delta and eta times, the input values and custom message. The logged values style are highly customizable.