Update app.py

app.py

@@ -4,6 +4,7 @@ from backtrack_sampler import BacktrackSampler, CreativeWritingStrategy
 from backtrack_sampler.provider.transformers_provider import TransformersProvider
 import torch
 import spaces
+import asyncio
 
 description = """## Compare Creative Writing: Custom Sampler vs. Backtrack Sampler with Creative Writing Strategy
 This is a demo of [Backtrack Sampler](https://github.com/Mihaiii/backtrack_sampler) using one of its algorithms named "Creative Writing Strategy".
@@ -41,8 +42,6 @@ def generate_responses(prompt, history):
 
     #already has special tokens
     inputs = tokenizer.encode(wrapped_prompt, add_special_tokens=False, return_tensors="pt").to("cuda")
-    def standard_task():
-        return model1.generate(inputs, max_length=2048, temperature=0.7)
 
     # Custom sampler task: loop over generator and collect outputs in a list
     async def custom_sampler_task():
@@ -51,8 +50,9 @@ def generate_responses(prompt, history):
         for token in generator:
             generated_list.append(token)
         return tokenizer.decode(generated_list, skip_special_tokens=True)
-    standard_output = standard_task()
-    custom_output = custom_sampler_task()
+        
+    custom_output = asyncio.run(custom_sampler_task())
+    standard_output = model1.generate(inputs, max_length=2048, temperature=0.7)
     # Decode standard output and remove the prompt from the generated response
     standard_response = tokenizer.decode(standard_output[0][len(inputs[0]):], skip_special_tokens=True)