Pritish23/flower-lora-weights

🌸 Flower LoRA Weights

LoRA fine-tuned weights for Stable Diffusion v1.5 trained on the Oxford-102 Flowers dataset.

Model Details

• Base model: Stable Diffusion v1.5
• Fine-tuning: LoRA (r=16, alpha=32)
• Dataset: Oxford-102 Flowers (1,020 images)
• Epochs: 10
• CLIP Score: 0.315

Usage

from diffusers import StableDiffusionPipeline
from peft import PeftModel
import torch

pipeline = StableDiffusionPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    torch_dtype=torch.float16
)
pipeline.unet = PeftModel.from_pretrained(
    pipeline.unet,
    "Pritish23/flower-lora-weights"
)
pipeline.unet = pipeline.unet.merge_and_unload()
pipeline = pipeline.to("cuda")

image = pipeline("a photo of a sunflower, vibrant colors").images[0]
image.save("sunflower.png")

Training Details

• Target modules: to_q, to_k, to_v, to_out
• Learning rate: 1e-4
• Batch size: 1
• Optimizer: AdamW

Demo

Try the live demo: Text-to-Image Generator Space

Author

Pritish Sharma — B.Sc. Data Science

• GitHub: Pritish-23
• LinkedIn: pritishsharma230805