Update app.py

app.py

@@ -17,7 +17,7 @@ class ModelActor:
         """
         self.model = AutoPipelineForText2Image.from_pretrained("stabilityai/sdxl-turbo")
 
-    def generate_image(self, prompt, prompt_name):
+    async def generate_image(self, prompt, prompt_name):
         """
         Generates an image based on the provided prompt.
         Parameters:
@@ -29,7 +29,7 @@ class ModelActor:
         start_time = time.time()
         process_id = os.getpid()
         try:
-            output = self.model(prompt=prompt, num_inference_steps=1, guidance_scale=0.0)
+            output = await self.model(prompt=prompt, num_inference_steps=1, guidance_scale=0.0)
             if isinstance(output.images, list) and len(output.images) > 0:
                 image = output.images[0]
                 buffered = BytesIO()
@@ -63,8 +63,7 @@ async def queue_api_calls(sentence_mapping, character_dict, selected_style):
     model_actors = [ModelActor.remote() for _ in range(num_actors)]
     tasks = [model_actors[i % num_actors].generate_image.remote(prompt, f"Prompt {paragraph_number}") for i, (paragraph_number, prompt) in enumerate(prompts)]
 
-    # Use asyncio.to_thread to call ray.get() synchronously in an async context
-    responses = await asyncio.gather(*[asyncio.to_thread(ray.get, task) for task in tasks])
+    responses = await asyncio.gather(*[ray.get(task) for task in tasks])
     images = {paragraph_number: response for (paragraph_number, _), response in zip(prompts, responses)}
     return images
 
@@ -91,7 +90,7 @@ gradio_interface = gr.Interface(
     fn=process_prompt,
     inputs=[gr.JSON(label="Sentence Mapping"), gr.JSON(label="Character Dict"), gr.Dropdown(["oil painting", "sketch", "watercolor"], label="Selected Style")],
     outputs="json"
-).queue(default_concurrency_limit=20)  # Adjust the concurrency limit as needed
+).queue(default_concurrency_limit=20)  # Set concurrency limit to match the number of model actors
 
 if __name__ == "__main__":
     gradio_interface.launch()