OpenAI — Jukebox [Attempt #2]
I tested Jukebox with my own music.
I am now taking my second attempt at using OpenAI’s Jukebox model. The first attempt seemed to be going well throughout layers of processing. The model generation rendered a few batches of low quality audio in 40mins-1hr. Once upsampling began, the model stopped midway through rendering audio with a “lack of sufficient space” popup.
I ran an additional block of code in my second attempt to boost my Google Colab storage. To be safe, I also reduced original sample time and the amount of music to be generated. I ran the script again and it yielded better results than the first attempt but still seemed lackluster. The model successfully got past a few more layers but still ultimately crashed.
Google Colab storage is a limiting factor that leaves me thinking of alternative methods of running my script.
I reinstalled all associated packages and libraries used to create the first two attempts via Google Colab.
Everything was set to go…but didn’t work.
I decided to prep my laptop to run this model locally. I have a Razer Blade 15 (2020) and it has a good GPU, RTX 2080. It is definitely worth a shot. I will let you know how it goes when I get pass the import errors. ^^
