lora_eval

Overview

lora_eval is a tool for evaluating LoRA (Low-Rank Adaptation) fine-tuned models using the automatic1111 Stable Diffusion API. It facilitates the comparison of different fine-tunes by generating and organizing output images.

Requirements

• A trained lora compatible with the automatic1111 API
• A running instance of the automatic1111 API server

Installation

1. Clone the lora_eval repository
2. Install dependencies: pip install -r requirements.txt
3. Ensure the automatic1111 API server is running

Usage

Run evaluate_lora.py with the necessary parameters to generate images:

 python3 lora_eval.py -l "christmas" --lora_strength -4 -2 0 2 4 -t "christmas" -url http://localhost:7777/sdapi/v1/txt2img

Run lora_anim.py to generate video:

mkdir v5
mkdir anim2
CUDA_VISIBLE_DEVICES=0 python3 lora_anim.py -s -2.3 -e 2.3 -l "happy" -p " __sdprompt__, __bg__ BREAK happy, __person__" -np "blur, blurry" -n 2 -sd 1 -m 120 -url "http://192.168.0.180:7777/sdapi/v1/txt2img" -f "anim2" -b 200