https://github.com/sumny/hpodash

Some preprocessing scripts for pre-collected HPO data
https://github.com/sumny/hpodash

Last synced: about 1 month ago
JSON representation

Some preprocessing scripts for pre-collected HPO data

• Host: GitHub
• URL: https://github.com/sumny/hpodash
• Owner: sumny
• License: mit
• Created: 2023-11-17T08:28:09.000Z (about 1 year ago)
• Default Branch: main
• Last Pushed: 2023-11-17T15:13:06.000Z (about 1 year ago)
• Last Synced: 2024-12-17T13:46:41.363Z (about 1 month ago)
• Language: Python
• Homepage:
• Size: 10.7 KB
• Stars: 2
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# hpodash

A collection of preprocessing scripts for getting existing pre-collected HPO data into a common long `csv` format.

This is for example useful to quickly visualize learning curves for all configurations of a given task.

## Notes:

* Preprocessing is usually not very memory efficient, and you will need at least 32GB of RAM and the resulting `csv` files can grow quite large - this should be fine because if you want this long `csv` format you will run the preprocessing only once and then use the `csv` files.

* The original data sources and preprocessed `csv` files are not included in this repository, but the scripts to obtain them.

* Scripts should work fine with Python 3.10+.

## Data sources considered so far:

* fcnet

* lcbench

* pd1

## Other data sources not considered so far:

* TaskSet (https://github.com/google-research/google-research/blob/master/task_set/README.md)

* TabZilla? (https://github.com/naszilla/tabzilla)

* TabRepo (https://github.com/autogluon/tabrepo)

* Revisiting Deep Learning Models for Tabular Data (https://github.com/yandex-research/tabular-dl-revisiting-models)

* Why Do Tree-Based Models Still Outperform Deep Learning on Tabular Data? (https://github.com/LeoGrin/tabular-benchmark)