HARRY07979/stable-diffusion-v1-5-openvino

OpenVINO™ Stable Diffusion v1.5

This is the OpenVINO™ version of runwayml/stable-diffusion-v1-5 (or stable-diffusion-v1-5/stable-diffusion-v1-5) . Optimized for CPUs, especially for Intels. Also, this version size is only about 5 GB.

HOW TO USE THIS MODEL

First, install needed packages:

pip install optimum[intel] diffusers transformers safetensors huggingface_hub torch numpy Pillow

Then, load model with OVStableDiffusionPipeline:

# Imports
from optimum.intel.openvino import OVStableDiffusionPipeline
from huggingface_hub import login
import torch
from pathlib import Path

# Configuration
MODEL_ID = "HARRY07979/stable-diffusion-v1-5-openvino"
DEVICE = "cpu"          # change to "gpu" if you have an Intel GPU
DTYPE = torch.float32 if DEVICE == "cpu" else torch.float16

# Load pipeline
pipe = OVStableDiffusionPipeline.from_pretrained(
    MODEL_ID,
    export=True,
    device=DEVICE,
    torch_dtype=DTYPE,
)

# Generation parameters
prompt = "A photorealistic portrait of Harry Potter, detailed, cinematic lighting"
negative_prompt = "low quality, blurry, distorted"
generator = torch.Generator().manual_seed(42)

# Generate image
output = pipe(
    prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=30,
    guidance_scale=7.5,
    generator=generator,
)

# Save result
output_dir = Path("generated")
output_dir.mkdir(parents=True, exist_ok=True)
output.images[0].save(output_dir / "harry_potter.png")
print(f"Image saved to {output_dir / 'harry_potter.png'}")

Contributing

If you are interested with this project, you can contribute me. This is an open project and I welcome all contributions. Feel free to open a Pull Request!

Thanks! 🥰🥰🥰🥰😆🤗