File size: 1,336 Bytes
e22ee3d
ded4997
5813abc
e10f7f0
e22ee3d
 
e789d51
e22ee3d
 
e10f7f0
5813abc
e22ee3d
 
e10f7f0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e789d51
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
---
title:  SD1.5 Interior LoRA
emoji: 🚀
colorFrom: indigo
colorTo: purple
sdk: gradio
sdk_version: 6.11.0
app_file: app.py
pinned: false
license: apache-2.0
tag: i
---

# LCM DreamShaper v7 — OpenVINO INT8

Fast CPU-based text-to-image generation using [Latent Consistency Models](https://latent-consistency-models.github.io/) with OpenVINO INT8 quantization.

## Models used
- [`OpenVINO/LCM_Dreamshaper_v7-int8-ov`](https://huggingface.co/OpenVINO/LCM_Dreamshaper_v7-int8-ov) — main pipeline
- [`rupeshs/taesd-ov`](https://huggingface.co/rupeshs/taesd-ov) — tiny autoencoder for fast decoding

## Setup

The Space expects the model folder `LCM_Dreamshaper_v7-int8-ov` to exist at the repo root.  
Use the download script below **once**, commit the weights, and the app will load from disk.

```python
import huggingface_hub as hf_hub

hf_hub.snapshot_download("OpenVINO/LCM_Dreamshaper_v7-int8-ov", local_dir="LCM_Dreamshaper_v7-int8-ov")
hf_hub.snapshot_download("rupeshs/taesd-ov", allow_patterns="vae_decoder/*", local_dir="LCM_Dreamshaper_v7-int8-ov")
hf_hub.snapshot_download("rupeshs/taesd-ov", allow_patterns="vae_encoder/*", local_dir="LCM_Dreamshaper_v7-int8-ov")
```

Or add a `download_models.py` script and call it from a Space startup script.

## Requirements

```
gradio>=4.44.0
openvino-genai
Pillow
numpy
```