search

EpistasisLab / Aliro

Aliro: AI-Driven Data Science
https://epistasislab.github.io/Aliro
GNU General Public License v3.0
225 stars 63 forks source link

Challenge associated with executing pickle files #648

Open HyunjunA opened 8 months ago

HyunjunA commented 8 months ago

Currently, running some pickle files trained on Aliro necessitates Python version 3.7.16 and sklearn version 0.24.2. Users will encounter issues testing models if their sklearn version is incompatible with the version in which the model's pickle file was created.

We provide the Colab script tailored for collaborators, such as doctors and nurses, enabling them to load pickle files from Aliro and test them on a testing set without the need for environment configuration. However, for example, specific model, like the GradientBoostingClassifier generate error messages due to sklearn version mismatches.

My solution employs the Aliro machine Docker container for loading pickle files trained on Aliro. Since this container is equipped with Python version 3.7.16 and sklearn version 0.24.2, it facilitates seamless pickle file loading and model testing. Nonetheless, this approach may prove challenging for users without development experience to utilize and follow.

HyunjunA commented 7 months ago

Surveying the application of Miguel's idea to the current Aliro system: Develop and integrate a function that allows users to execute pickle files within Aliro.