Update app.py

app.py

@@ -5,10 +5,8 @@ torch.backends.cudnn.allow_tf32 = True
 import gradio as gr
 import numpy as np
 import random
-import spaces
 import time
 from diffusers import DiffusionPipeline, AutoencoderTiny
-from diffusers.models.attention_processor import AttnProcessor2_0
 from custom_pipeline import FluxWithCFGPipeline
 import asyncio
 
@@ -51,8 +49,9 @@ if hasattr(pipe, "transformer") and torch.cuda.is_available():
 
 torch.cuda.empty_cache()
 
-# Inference function
-@spaces.GPU(duration=25)
+
+
+# Inference function (async)
 async def generate_image(
     prompt,
     seed=24,
@@ -119,8 +118,8 @@ async def generate_image(
             static_latents_out, height, width, "pil"
         )
 
-    # Graph-based generation function
-    async def generate_with_graph(
+    # Graph-based generation function (synchronous)
+    def generate_with_graph(
         latents,
         prompt_embeds,
         pooled_prompt_embeds,
@@ -137,8 +136,7 @@ async def generate_image(
         g.replay()
         return static_output
 
-    # Only generate the last image in the sequence
-    img = await pipe.generate_images(
+    img = pipe.generate_images(
         prompt=prompt,
         width=width,
         height=height,