Update app.py

app.py

@@ -4,21 +4,20 @@ from generate_prompts import generate_prompt
 from diffusers import AutoPipelineForText2Image
 from io import BytesIO
 import gradio as gr
-from multiprocessing import Pool, current_process
+from concurrent.futures import ThreadPoolExecutor
 
-def load_model():
-    print(f"Loading model in process {current_process().name}")
-    model = AutoPipelineForText2Image.from_pretrained("stabilityai/sdxl-turbo")
-    print(f"Model loaded in process {current_process().name}")
-    return model
+# Load the model once outside of the function
+print("Loading the model...")
+model = AutoPipelineForText2Image.from_pretrained("stabilityai/sdxl-turbo")
+print("Model loaded successfully.")
 
 def generate_image(prompt, prompt_name):
     try:
-        model = load_model()
         print(f"Generating response for {prompt_name} with prompt: {prompt}")
         output = model(prompt=prompt, num_inference_steps=1, guidance_scale=0.0)
         print(f"Output for {prompt_name}: {output}")
 
+        # Check if the model returned images
         if isinstance(output.images, list) and len(output.images) > 0:
             image = output.images[0]
             buffered = BytesIO()
@@ -40,6 +39,7 @@ async def queue_api_calls(sentence_mapping, character_dict, selected_style):
     print(f"queue_api_calls invoked with sentence_mapping: {sentence_mapping}, character_dict: {character_dict}, selected_style: {selected_style}")
     prompts = []
 
+    # Generate prompts for each paragraph
     for paragraph_number, sentences in sentence_mapping.items():
         combined_sentence = " ".join(sentences)
         print(f"combined_sentence for paragraph {paragraph_number}: {combined_sentence}")
@@ -47,10 +47,15 @@ async def queue_api_calls(sentence_mapping, character_dict, selected_style):
         prompts.append((paragraph_number, prompt))
         print(f"Generated prompt for paragraph {paragraph_number}: {prompt}")
 
-    with Pool() as pool:
-        tasks = [(prompt, f"Prompt {paragraph_number}") for paragraph_number, prompt in prompts]
+    # Set max_workers to the total number of prompts
+    max_workers = len(prompts)
+
+    # Generate images for each prompt in parallel using threading
+    with ThreadPoolExecutor(max_workers=max_workers) as executor:
+        loop = asyncio.get_running_loop()
+        tasks = [loop.run_in_executor(executor, generate_image, prompt, f"Prompt {paragraph_number}") for paragraph_number, prompt in prompts]
         print("Tasks created for image generation.")
-        responses = pool.starmap(generate_image, tasks)
+        responses = await asyncio.gather(*tasks)
         print("Responses received from image generation tasks.")
 
     images = {paragraph_number: response for (paragraph_number, _), response in zip(prompts, responses)}
@@ -60,16 +65,20 @@ async def queue_api_calls(sentence_mapping, character_dict, selected_style):
 def process_prompt(sentence_mapping, character_dict, selected_style):
     print(f"process_prompt called with sentence_mapping: {sentence_mapping}, character_dict: {character_dict}, selected_style: {selected_style}")
     try:
+        # See if there is a loop already running. If there is, reuse it.
         loop = asyncio.get_running_loop()
     except RuntimeError:
+        # Create new event loop if one is not running
         loop = asyncio.new_event_loop()
         asyncio.set_event_loop(loop)
     print("Event loop created.")
 
+    # This sends the prompts to function that sets up the async calls. Once all the calls to the API complete, it returns a list of the gr.Textbox with value= set.
     cmpt_return = loop.run_until_complete(queue_api_calls(sentence_mapping, character_dict, selected_style))
     print(f"process_prompt completed with return value: {cmpt_return}")
     return cmpt_return
 
+# Gradio interface with high concurrency limit
 gradio_interface = gr.Interface(
     fn=process_prompt,
     inputs=[