Additional notes

!!!This project is not actively maintained. It is for archival purposes only and mustn't be hosted on any public server!!!

• place a huggingface_token.txt in the ai/-folder and add a Huggingface access token to download the Llama-2 model (might need to apply for access)

• all of the images used to train the custom image model and their descriptions are in the model training images folder. The model was trained using the following settings:

  Name	Value
  Epochs	8
  Learning Rate	1e-05
  Learning Rate Scheduler	cosine
  Learning Rate Warmup %	10
  Batch Size	2
  Buckets	Yes
  Optimizer	AdamW8bit
  Base Model	SDXL-Base-1.0
  Precision	bf16

• AI requirements can be installed with pip install -r .\requirements.txt --extra-index-url https://download.pytorch.org/whl/cu118

• Frontend requirements can be installed as usual using npm or yarn