Fine-tuning riffusion/riffusion-model-v1 on a custom dataset in Google Colab

[yukiarimo]

Hello. I have a dataset in a folder with songs and textual descriptions for them like this (split into 5 seconds of audio):

ls ./dataset

1.wav
1.txt
2.wav
2.txt

Can you please show me how to fine-tune your model on my dataset for 512x512 images/MELs (with code/notebook, if possible) in Google Colab on A100 or L4?