mlmnemonist 0.0.16

A light-weight framework to store the progress you madeon your ML operations with the ability to smartly cache your modelsand retrieve it even when your session crashes.

ML-Mnemonist is an open-source lightweight framework for simple ML operations and it pairs up easily with free online cloud services that do not support session backup and recovery because the framework implements these in the hardware level. To that end, some of the tutorials here are purposefully created to work with Google Colab. <p align=”center”> <img src=”/figures/logo.png”/> </p>

Check out more information on the GitHub <a href=”https://github.com/HamidrezaKmK/ML-Mnemonist”>repository</a>.

Change Log

0.0.15 (09.10.2022)

• Internal functionalities of Experiment Runner changed

• Creating new experiments with the same name overwrites the previous experiment with the same name rather than creating new ones

• The caching scheme for meta parameters had a bug which is fixed

• The Tutorial is finalized

0.0.14 (19.7.2022)

• Add exception handler to hyper experiments for conflicting configs

0.0.13 (01.7.2022)

• Bug removal in configurations of the factory

• Bug removal in readme

0.0.12 (01.7.2022)

• Biggest update yet!

• Introduce HyperExperiments!

• When retrieving using factory, the pipelines will also be handled

• Caching system updated, META_DATA cache added

0.0.11 (01.7.2022)

• Grid search can now continue on the runner’s last interrupt

0.0.10 (01.7.2022)

• With preprocess added to gridsearch

• merge_cfg added to ExperimentRunner

0.0.9 (01.7.2022)

• Grid search bug removal

• Implement two types of grid search: one using a palette config and another using a directory containing all the config files

• Retrieve added to ExperimentRunnerFactory

• Updated naming scheme for experiments such that each experiment has a description file and their name contains the date of today

0.0.8 (01.7.2022)

• Bug removal in validation tools

0.0.7 (01.7.2022)

• Add expand_cfg for hyperparameter tuning

• Add gridsearch for hyperparameter tuning

• Change cache token scheme

0.0.6 (29.6.2022)

• Bug removal in CACHE.RESET() where directories are included in the logs

• Runner pipeline cloning option added

• preprocess does keep the pipeline by default

0.0.5 (24.6.2022)

• Add reload_cfg before each call of runner.run

• Add reload_cfg before each call of runner.preprocess

• Add runner.export_logs() method to get a zip file from all logfiles in the experiment directory

0.0.4 (23.6.2022)

• Add more logs for verbose runners

• Add CACHE.LOGS_DIR to save logs such as tensorboard

• Add maximum cache limit that can be manipulated using mlm.MAX_CACHE_SIZE

0.0.3 (23.6.2022)

• README updated

• full support without configurations

• secret_root added to factory and reveal_true_path

0.0.2 (23.6.2022)

• Bug removal of some initial problems with the setup

0.0.1 (23.6.2022)

• First release