decrease samples

app.py

@@ -223,7 +223,7 @@ def llm_call(question_prompt, model_name,
     top_p=1, n_samples=64, stop=None):
     if HUGGINGFACE:
         model_inputs = hug_tokenizer([question_prompt], return_tensors="pt").to('cuda')
-        generated_ids = hug_model.generate(**model_inputs, max_length=1400, temperature=1, num_return_sequences=64, do_sample=True)
+        generated_ids = hug_model.generate(**model_inputs, max_length=1400, temperature=1, num_return_sequences=32, do_sample=True)
         responses = hug_tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
         codes = []
         for response in responses: