Update app.py

app.py

@@ -20,9 +20,6 @@ user_prompt = '<|user|>\n'
 assistant_prompt = '<|assistant|>\n'
 prompt_suffix = "<|end|>\n"
 
-params = og.GeneratorParams(model)
-params.set_search_options(max_length=4096)    
-
 # Inference function
 def ask_phi(image, question):
     prompt = f"{user_prompt}<|image_1|>\n{question}{prompt_suffix}{assistant_prompt}"
@@ -30,6 +27,8 @@ def ask_phi(image, question):
     inputs = processor(prompt=prompt, images=images, return_tensors="pt")
     
     start_time = time.time()
+    params = og.GeneratorParams(model)
+    params.set_search_options(max_length=4096)    
     params.set_inputs(inputs)
     generator = og.Generator(model, params)
     print(f"Generator setup took {time.time() - start_time} seconds")
@@ -41,7 +40,6 @@ def ask_phi(image, question):
         new_token = generator.get_next_tokens()[0]
         response += tokenizer_stream.decode(new_token)
 
-    del generator
     return response
 
 # Gradio Interface