Update app.py

app.py

@@ -3,25 +3,20 @@ import numpy as np
 import random
 from diffusers import DiffusionPipeline
 import torch
-
-# Check for GPU availability and set the device accordingly
-device = "cuda" if torch.cuda.is_available() else "cpu"
-
-# Load the diffusion pipeline based on the availability of GPU
-if device == "cuda":
-    pipe = DiffusionPipeline.from_pretrained("stabilityai/sdxl-turbo", torch_dtype=torch.float16, variant="fp16", use_safetensors=True)
-    pipe.enable_xformers_memory_efficient_attention()
-    pipe = pipe.to(device)
-else: 
-    pipe = DiffusionPipeline.from_pretrained("stabilityai/sdxl-turbo", use_safetensors=True)
-    pipe = pipe.to(device)
+import spaces
 
 # Constants
 MAX_SEED = np.iinfo(np.int32).max
 MAX_IMAGE_SIZE = 1024
 
-# Inference function
+# Load the diffusion pipeline without immediately moving it to GPU
+pipe = DiffusionPipeline.from_pretrained("stabilityai/sdxl-turbo", torch_dtype=torch.float16, use_safetensors=True)
+
+@spaces.GPU
 def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps):
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+    pipe.to(device)
+    
     if randomize_seed:
         seed = random.randint(0, MAX_SEED)
         
@@ -55,7 +50,7 @@ css="""
 """
 
 # Determine the power device (GPU or CPU) for display purposes
-power_device = "GPU" if device == "cuda" else "CPU"
+power_device = "GPU" if torch.cuda.is_available() else "CPU"
 
 # Gradio UI setup
 with gr.Blocks(css=css) as demo: