metadata

license: openrail
language:
  - en
base_model:
  - stabilityai/stable-diffusion-3.5-medium
pipeline_tag: text-to-image

Stable Diffusion 3.5 Medium Turbo (SD3.5M Turbo)

"All that has passed is but a prologue，what you hold in your heart will one day echo back."

Project Overview

Stable Diffusion 3.5 Medium Turbo (SD3.5M Turbo) is a high-performance text-to-image model distilled from StabilityAI's stable-diffusion-3.5-medium. This model emphasizes stability and efficiency, making it suitable for a wide range of art styles and creative expression scenarios.

Model Features

🚀 Turbo Performance: Faster generation speeds, ideal for multitasking and high-demand scenarios.
🎨 Versatile Styles: Supports a wide range of styles, from photorealistic to abstract art.
🖼️ High-Resolution Outputs: Produces images with exceptional clarity and intricate details.
⚙️ Easy to Extend: Integrated with LoRA technology, making it easier for users to customize and experiment.

How to Use

Download the Model
Download the latest versions of the model’s ckpt and LoRA files from the following links:
- SD3.5M Checkpoint (Updated on 2024/12/16)
- SD3.5M LoRA (Updated on 2024/12/15)
Environment Setup
Ensure that your environment meets the following requirements:
- Python 3.8+
- PyTorch 2.0+
- Required libraries such as diffusers
Model Loading
Load and use the model following the detailed instructions provided in the repository.

Example Output

import torch
from diffusers import StableDiffusion3Pipeline

pipe = StableDiffusion3Pipeline.from_pretrained("tensorart/stable-diffusion-3.5-medium-turbo", torch_dtype=torch.float16,)
                                                
pipe = pipe.to("cuda")


image = pipe(
   "A beautiful bald girl with silver and white futuristic metal face jewelry, her full body made of intricately carved liquid glass in the style of Tadashi, the complexity master of cyberpunk, in the style of James Jean and Peter Mohrbacher. This concept design is trending on Artstation, with sharp focus, studio-quality photography, and highly detailed, intricate details.",
   num_inference_steps=8,
   guidance_scale=1.5,
   height=1024,
   width=768 
).images[0]

image.save("./test4-2.webp")

Using lora：

import torch
from diffusers import StableDiffusion3Pipeline
import numpy as np
from safetensors.torch import load_file

pipe = StableDiffusion3Pipeline.from_pretrained("tensorart/stable-diffusion-3.5-medium-turbo", torch_dtype=torch.float16,)
                                                
pipe = pipe.to("cuda")

lora_weights_dir = 'tensorart/stable-diffusion-3.5-medium-turbo/lora_sd3.5m_turbo_8steps.safetensors'
pcm_lora_weight = load_file(lora_weights_dir)

alpha = 1.0
pcm_lora_weight = {
   key: value * np.sqrt(alpha) for key, value in pcm_lora_weight.items()
}
pipe.load_lora_weights(lora_weights_dir)
pipe = pipe.to("cuda")

image = pipe(
   "A beautiful bald girl with silver and white futuristic metal face jewelry, her full body made of intricately carved liquid glass in the style of Tadashi, the complexity master of cyberpunk, in the style of James Jean and Peter Mohrbacher. This concept design is trending on Artstation, with sharp focus, studio-quality photography, and highly detailed, intricate details.",
   num_inference_steps=8,
   guidance_scale=1.5,
   height=1024,
   width=768 
).images[0]
image.save("./test1.webp")