update

utils/model.py

@@ -56,7 +56,7 @@ class Model(torch.nn.Module):
             sampling_params = SamplingParams(
                 temperature=temp,
                 max_tokens=max_length,
-                top_p=0.95 if do_sample else 1.0,
+                #top_p=0.95 if do_sample else 1.0,
                 stop_token_ids=[self.tokenizer.eos_token_id]
             )
             outputs = self.llm.generate(content_list, sampling_params)