How to train your own dataset? - Githubissues

LambdaLabsML / examples

Deep Learning Examples

MIT License

805 stars 103 forks source link

How to train your own dataset? #20

Closed zhenzi0322 closed 1 year ago

zhenzi0322 commented 1 year ago

How to train your own dataset?

pfdamasceno commented 1 year ago

Check out https://github.com/justinpinkney/stable-diffusion/blob/main/ldm/data/simple.py and either use or modify FolderData for your purposes. Make sure to edit the yaml file as well to use FolderData instead of hf_dataset.

justinpinkney commented 1 year ago

Also see this: https://github.com/LambdaLabsML/examples/issues/16#issuecomment-1274544700

You can make your own directory and a jsonl file with metadata according to huggingfaces' specification then directly use that.