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

[adwaykanhere]

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]

✅ Deploy Preview for dfdn canceled.

Name	Link
🔨 Latest commit	5df413e
🔍 Latest deploy log	https://app.netlify.com/sites/dfdn/deploys/628288953cd88900084f4f52

[PSSF23]

You need to optimize the requirements and dependencies. Why removing the versions?

Say if script A imports package I, and script B imports script A, then script B should not import package I. Remove the duplications to make the code more robust.

Additionally, I don't see why there's an fsd18k and an fsd18ktuning. Leave the updated one and name it fsd18k.

[PSSF23]

fsdk18preprocess is not necessary as a separate file. Put the function inside toolbox.

[adwaykanhere]

all changes updated

[PSSF23]

I don't understand why you downgraded opencv-python and use a different format for torch packages (checks not passed). Also there is no sklearn pacakge name in pip.

[adwaykanhere]

Sorry, the changes have been fixed.

[PSSF23]

Did you use botorch at all?

[adwaykanhere]

Botorch is a requirement for Ax. Reference