Added Bayesian Hyperparameter tuning for all DN on Audition tasks (Issue #31)

adwaykanhere commented 2 years ago

This PR will address part of Issue #31 and have the following changes:

Closed adding Bayesian hyperparameter tuning of the Audition data using Pytorch's Ax library and Raytune.
Fixed adding a separate file- fsdk18preprocess.py for all preprocessing steps required to run DN benchmarks on audition. This is necessary to run fsdk18.py and fsdk18tuning.py scripts.
Ensured that the ratio of sample sizes for audition is consistently 2:1:1 for all the class combinations when running benchmarks with and without tuning.
Resolved to remove any unnecessary functions in the audition toolbox

Library conflicts: Raytune needs to be version 1.11.0 and not later

netlify[bot] commented 2 years ago

Name	Link
Latest commit	5df413e06ce33c6be5d005e6d1141de9fcd45cb4
Latest deploy log	https://app.netlify.com/sites/dfdn/deploys/628288953cd88900084f4f52

adwaykanhere commented 2 years ago

all changes updated

adwaykanhere commented 2 years ago

Sorry, the changes have been fixed.

adwaykanhere commented 2 years ago

Botorch is a requirement for Ax. Reference

neurodata / df-dn-paper