search

Marigoldwu / A-Unified-Framework-for-Deep-Attribute-Graph-Clustering

This project is a scalable unified framework for deep graph clustering.
https://www.marigold.website/readArticle?workId=145&author=Marigold&authorId=1000001
MIT License
90 stars 12 forks source link

[Enhancement] Log loss values when training #6

Open habaneraa opened 12 months ago

habaneraa commented 12 months ago

When pre-training/training the GNNs, it would be better if the loss values per step are logged. Sometimes the metric scores may not change despite the decreasing loss, and the loss values can effectively indicate the progress of self-supervised learning.

A possible solution is to add a new key-value argument (e.g. loss=f"{loss.detach().item()}") when calling logger.info(get_format_variables()) in every training script.

Marigoldwu commented 12 months ago

Sounds great! Thank you for your valuable suggestions and we will incorporate them in the upcoming versions. Thanks again for using this framework!

habaneraa commented 12 months ago

This project is helpful. I'm glad to help with building a more robust and user-friendly framework for DAGC!