Edna Naya 1.0
SDXL LoRA adapter (r=16, alpha=32, dropout 0.05, fp16) fine-tuned for photorealistic high-definition image generation. Trained on 18 curated real-photo datasets: portraits/faces, Unsplash & Flickr photography, landscapes, studio product shots, and cityscapes (resolution-filtered, deduplicated, 1024px). Part of project Edna Naya 1.0 by Lobus22.
Usage
from diffusers import StableDiffusionXLPipeline
import torch
pipe = StableDiffusionXLPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16)
pipe.load_lora_weights("Lobus22/Edna-Naya-1.0")
pipe = pipe.to("cuda")
Prompt style: "a realistic photograph of ..., high detail, natural lighting".
Trained on Kaggle 2x T4: LR probe then cosine schedule, val loss every
250 steps, early stopping (patience 3), best checkpoint pushed.
Sample outputs in samples/.
- Downloads last month
- -
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for Lobus22/Edna-Naya-1.0
Base model
stabilityai/stable-diffusion-xl-base-1.0