About

The small sd model is a streamlined and optimized version of a text-to-image generator, designed to produce realistic visuals from natural language descriptions. Distilled from a high-performing base model, it stands out for its speed and efficiency while maintaining satisfactory image quality. Ideal for applications requiring quick renders or limited resources, it is particularly well-suited for content creators, digital artists, or automated tools where speed is a priority. Its lightweight nature makes it accessible without high-end hardware, while still delivering consistent results for a variety of prompts. This model serves as a practical alternative to heavier versions, without entirely compromising visual quality.

Documentation

license: creativeml-openrail-m base_model: SG161222/Realistic_Vision_V4.0 datasets:

recastai/LAION-art-EN-improved-captions tags:
stable-diffusion
stable-diffusion-diffusers
text-to-image
diffusers inference: true

Text-to-image Distillation

This pipeline was distilled from SG161222/Realistic_Vision_V4.0 on a Subset of recastai/LAION-art-EN-improved-captions dataset. Below are some example images generated with the finetuned pipeline using small-sd model.

val_imgs_grid

This Pipeline is based upon the paper. Training Code can be found here.

Pipeline usage

You can use the pipeline like so:

Python

from diffusers import DiffusionPipeline
import torch

pipeline = DiffusionPipeline.from_pretrained("segmind/small-sd", torch_dtype=torch.float16)
prompt = "Portrait of a pretty girl"
image = pipeline(prompt).images[0]
image.save("my_image.png")

Training info

These are the key hyperparameters used during training:

Steps: 95000
Learning rate: 1e-4
Batch size: 32
Gradient accumulation steps: 4
Image resolution: 512
Mixed-precision: fp16

Capabilities & Tags

diffusersstable-diffusionstable-diffusion-diffuserstext-to-imageendpoints_compatible

Links & Resources