Spaces:
Running
Running
Commit
•
4fa2b26
1
Parent(s):
565579e
Update app.py
Browse files
app.py
CHANGED
@@ -1,16 +1,33 @@
|
|
|
|
1 |
import torch
|
2 |
from diffusers import StableDiffusionXLPipeline, UNet2DConditionModel, EulerDiscreteScheduler
|
3 |
from huggingface_hub import hf_hub_download
|
4 |
from safetensors.torch import load_file
|
5 |
|
|
|
6 |
base = "stabilityai/stable-diffusion-xl-base-1.0"
|
7 |
repo = "ByteDance/SDXL-Lightning"
|
8 |
-
ckpt = "sdxl_lightning_4step_unet.safetensors"
|
9 |
|
10 |
-
# Load model
|
11 |
-
|
12 |
-
|
13 |
-
|
|
|
|
|
|
|
|
|
14 |
|
15 |
-
#
|
16 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
import torch
|
3 |
from diffusers import StableDiffusionXLPipeline, UNet2DConditionModel, EulerDiscreteScheduler
|
4 |
from huggingface_hub import hf_hub_download
|
5 |
from safetensors.torch import load_file
|
6 |
|
7 |
+
# Model Path/Repo Information
|
8 |
base = "stabilityai/stable-diffusion-xl-base-1.0"
|
9 |
repo = "ByteDance/SDXL-Lightning"
|
10 |
+
ckpt = "sdxl_lightning_4step_unet.safetensors"
|
11 |
|
12 |
+
# Load model (Executed only once for efficiency)
|
13 |
+
@st.cache_resource
|
14 |
+
def load_sdxl_pipeline():
|
15 |
+
unet = UNet2DConditionModel.from_config(base, subfolder="unet").to("cuda", torch.float16)
|
16 |
+
unet.load_state_dict(load_file(hf_hub_download(repo, ckpt), device="cuda"))
|
17 |
+
pipe = StableDiffusionXLPipeline.from_pretrained(base, unet=unet, torch_dtype=torch.float16, variant="fp16").to("cuda")
|
18 |
+
pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
|
19 |
+
return pipe
|
20 |
|
21 |
+
# Streamlit UI
|
22 |
+
st.title("Stable Diffusion XL Image Generation")
|
23 |
+
prompt = st.text_input("Enter your image prompt:")
|
24 |
+
|
25 |
+
if st.button("Generate Image"):
|
26 |
+
if not prompt:
|
27 |
+
st.warning("Please enter a prompt.")
|
28 |
+
else:
|
29 |
+
pipe = load_sdxl_pipeline() # Load the pipeline from cache
|
30 |
+
with torch.no_grad():
|
31 |
+
image = pipe(prompt).images[0]
|
32 |
+
|
33 |
+
st.image(image)
|