|
--- |
|
library_name: diffusers |
|
base_model: segmind/Segmind-Vega |
|
tags: |
|
- lora |
|
- text-to-image |
|
license: apache-2.0 |
|
inference: false |
|
--- |
|
# Segmind-VegaRT - Latent Consistency Model (LCM) LoRA of Segmind-Vega |
|
|
|
Try real-time inference here **[VegaRT demo⚡](https://www.segmind.com/segmind-vega-rt)** |
|
|
|
API for **[Segmind-VegaRT](https://www.segmind.com/models/segmind-vega-rt-v1/api)** |
|
|
|
<video controls autoplay src="https://cdn-uploads.huggingface.co/production/uploads/62039c2d91d53938a643317d/WacXd5DqP5hx8iEGTPt16.mp4"></video> |
|
|
|
Segmind-VegaRT a distilled consistency adapter for [Segmind-Vega](https://huggingface.co/segmind/Segmind-Vega) that allows |
|
to reduce the number of inference steps to only between **2 - 8 steps**. |
|
|
|
Latent Consistency Model (LCM) LoRA was proposed in [LCM-LoRA: A universal Stable-Diffusion Acceleration Module](https://arxiv.org/abs/2311.05556) |
|
by *Simian Luo, Yiqin Tan, Suraj Patil, Daniel Gu et al.* |
|
|
|
# Image comparison (Segmind-VegaRT vs SDXL-Turbo) |
|
|
|
 |
|
|
|
 |
|
|
|
 |
|
|
|
# Speed comparison (Segmind-VegaRT vs SDXL-Turbo) on A100 80GB |
|
|
|
 |
|
|
|
| Model | Params / M | |
|
|----------------------------------------------------------------------------|------------| |
|
| [lcm-lora-sdv1-5](https://huggingface.co/latent-consistency/lcm-lora-sdv1-5) | 67.5 | |
|
| [**Segmind-VegaRT**](https://huggingface.co/segmind/Segmind-VegaRT) | **119** | |
|
| [lcm-lora-sdxl](https://huggingface.co/latent-consistency/lcm-lora-sdxl) | 197 | |
|
|
|
## Usage |
|
|
|
LCM-LoRA is supported in 🤗 Hugging Face Diffusers library from version v0.23.0 onwards. To run the model, first |
|
install the latest version of the Diffusers library as well as `peft`, `accelerate` and `transformers`. |
|
audio dataset from the Hugging Face Hub: |
|
|
|
```bash |
|
pip install --upgrade pip |
|
pip install --upgrade diffusers transformers accelerate peft |
|
``` |
|
|
|
### Text-to-Image |
|
|
|
Let's load the base model `segmind/Segmind-Vega` first. Next, the scheduler needs to be changed to [`LCMScheduler`](https://huggingface.co/docs/diffusers/v0.22.3/en/api/schedulers/lcm#diffusers.LCMScheduler) and we can reduce the number of inference steps to just 2 to 8 steps. |
|
Please make sure to either disable `guidance_scale` or use values between 1.0 and 2.0. |
|
|
|
```python |
|
import torch |
|
from diffusers import LCMScheduler, AutoPipelineForText2Image |
|
|
|
model_id = "segmind/Segmind-Vega" |
|
adapter_id = "segmind/Segmind-VegaRT" |
|
|
|
pipe = AutoPipelineForText2Image.from_pretrained(model_id, torch_dtype=torch.float16, variant="fp16") |
|
pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config) |
|
pipe.to("cuda") |
|
|
|
# load and fuse lcm lora |
|
pipe.load_lora_weights(adapter_id) |
|
pipe.fuse_lora() |
|
|
|
|
|
prompt = "Self-portrait oil painting, a beautiful cyborg with golden hair, 8k" |
|
|
|
# disable guidance_scale by passing 0 |
|
image = pipe(prompt=prompt, num_inference_steps=4, guidance_scale=0).images[0] |
|
``` |
