revise parameters for advanced trainer

replicate / lora-training

LoRA training model packaged with Cog

MIT License

110 stars 29 forks source link

Open anotherjesse opened 1 year ago

anotherjesse commented 1 year ago

[ ] ensure descriptions are useful/accurate. example: class_data mentions class_prompt and num_class_images that don't exist
[ ] I've not seen other models expose device as an option... should we in lora-advanced-trainer?
[ ] continue_inversion - unless we expose sending in existing tensor(s) - does this make sense?
[ ] perform_inversion / train_text_encoder -- do both of these matter if we only return the last step_1000.safetensors?

cloneofsimo commented 1 year ago

Checked all the stuff you mentioned. train_text_encoder must be on in current implementation, but I'm going to update it in the future. #5