Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import torch
|
3 |
+
import os
|
4 |
+
from diffusers import StableDiffusion3Pipeline
|
5 |
+
from safetensors.torch import load_file
|
6 |
+
|
7 |
+
# Access HF_TOKEN from environment variables
|
8 |
+
hf_token = os.getenv("HF_TOKEN")
|
9 |
+
|
10 |
+
# Specify the pre-trained model ID
|
11 |
+
model_id = "stabilityai/stable-diffusion-3.5-large"
|
12 |
+
|
13 |
+
# Lazy pipeline initialization
|
14 |
+
pipeline = None
|
15 |
+
|
16 |
+
# Function for image generation
|
17 |
+
@gr.GPU(duration=65)
|
18 |
+
def generate_image(prompt): # Remove lora_file input
|
19 |
+
global pipeline
|
20 |
+
if pipeline is None:
|
21 |
+
try:
|
22 |
+
pipeline = StableDiffusion3Pipeline.from_pretrained(
|
23 |
+
model_id,
|
24 |
+
use_auth_token=hf_token,
|
25 |
+
torch_dtype=torch.float16,
|
26 |
+
cache_dir="./model_cache"
|
27 |
+
)
|
28 |
+
except Exception as e:
|
29 |
+
print(f"Error loading from cache: {e}")
|
30 |
+
pipeline = StableDiffusion3Pipeline.from_pretrained(
|
31 |
+
model_id,
|
32 |
+
use_auth_token=hf_token,
|
33 |
+
torch_dtype=torch.float16,
|
34 |
+
local_files_only=False
|
35 |
+
)
|
36 |
+
pipeline.enable_model_cpu_offload()
|
37 |
+
pipeline.enable_attention_slicing()
|
38 |
+
|
39 |
+
# Load and apply LoRA (file is already in the Space)
|
40 |
+
lora_filename = "lora_trained_model.safetensors" # Name of your LoRA file
|
41 |
+
lora_path = os.path.join("./", lora_filename) # Construct the path
|
42 |
+
print(f"Loading LoRA from: {lora_path}")
|
43 |
+
|
44 |
+
try:
|
45 |
+
if os.path.exists(lora_path): # check if the file exists
|
46 |
+
lora_weights = load_file(lora_path)
|
47 |
+
text_encoder = pipeline.text_encoder
|
48 |
+
text_encoder.load_state_dict(lora_weights, strict=False)
|
49 |
+
else:
|
50 |
+
return f"Error: LoRA file not found at {lora_path}"
|
51 |
+
except Exception as e:
|
52 |
+
return f"Error loading LoRA: {e}"
|
53 |
+
|
54 |
+
try:
|
55 |
+
image = pipeline(prompt).images[0]
|
56 |
+
return image
|
57 |
+
except Exception as e:
|
58 |
+
return f"Error generating image: {e}"
|
59 |
+
|
60 |
+
|
61 |
+
# Create the Gradio interface (remove lora_upload)
|
62 |
+
with gr.Blocks() as demo:
|
63 |
+
prompt_input = gr.Textbox(label="Prompt")
|
64 |
+
image_output = gr.Image(label="Generated Image")
|
65 |
+
generate_button = gr.Button("Generate")
|
66 |
+
|
67 |
+
generate_button.click(
|
68 |
+
fn=generate_image,
|
69 |
+
inputs=prompt_input, # Only prompt input now
|
70 |
+
outputs=image_output,
|
71 |
+
)
|
72 |
+
|
73 |
+
demo.launch()
|