Added base setup for training t2i adapters. Currently untested, saw something else shiny i wanted to finish sirst. Added content_or_style to the training config. It defaults to balanced, which is standard uniform time step sampling. If style or content is passed, it will use cubic sampling for timesteps to favor timesteps that are beneficial for training them. for style, favor later timesteps. For content, favor earlier timesteps.

This commit is contained in:
Jaret Burkett
2023-09-16 08:30:38 -06:00
parent 17e4fe40d7
commit 27f343fc08
8 changed files with 314 additions and 84 deletions

View File

@@ -1,7 +1,7 @@
torch
torchvision
safetensors
git+https://github.com/huggingface/diffusers.git
diffusers==0.21.1
transformers
lycoris_lora
flatten_json