Train a custom 4-bar models

allenhung1025 / LoopTest

Official repo of ISMIR-21 publication, “A Benchmarking Initiative for Audio-domain Music Generation using the FreeSound Loop Dataset”.

MIT License

77 stars 12 forks source link

Train a custom 4-bar models #4

Open littlebeanhp opened 1 year ago

littlebeanhp commented 1 year ago

Hello, thank you for this impressive work. It gave me much inspiration!

I want to train my own 4-bar model. As I dont have much music domain-knowledge, my plan is finetuning the pretrained model on a specific genre, with really low Learning rate. Can you provide me the step of resuming train step with four-bar model? Like how many seconds should we trim the datapoint, and so on...

Thank you so much