https://github.com/kayoyin/prodigymodels

Pretrained models for Project Prodigy
https://github.com/kayoyin/prodigymodels

Last synced: 8 days ago
JSON representation

Pretrained models for Project Prodigy

• Host: GitHub
• URL: https://github.com/kayoyin/prodigymodels
• Owner: kayoyin
• Created: 2019-01-18T11:53:27.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2019-01-19T13:41:52.000Z (almost 6 years ago)
• Last Synced: 2023-10-20T20:55:32.800Z (about 1 year ago)
• Homepage:
• Size: 6.84 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# ProdigyModels

Collection of some of the models trained by [Project Prodigy](https://github.com/RayDiab/Prodigy)

Make sure to have [Git LFS](https://git-lfs.github.com/) installed before cloning the repository.

- `jazzy.xml`: trained on two short pieces of Bach, gets stuck into a cycle during prediction, 512 hidden units. `rho = 5`

- `easybach.xml`: trained on a collection of Bach's works with simple combination of notes, 256 hidden units. Also the model used in the GUI in Project Prodigy. `rho = 20`

- `512easy.xml`: trained on the same collection of Bach's works as above, with 512 hidden units. `rho = 5`

- `river.xml`: trained on Yiruma's "River flows into you", with 256 hidden units. `rho = 10`

- `sixbach.xml`: trained on six long Bach pieces, with 256 hidden units. Gives chaotic compositions. `rho = 20`

More experimenting with parameters should be done but so far, the best-performing models seem to be the ones with 512 hidden units with `rho` between 5 and 10.