top-p, temperature 조정

app.py

@@ -16,7 +16,7 @@ tokenizer = AutoTokenizer.from_pretrained(model_name)
 #     device="cuda:0" if torch.cuda.is_available() else 'cpu'
 # )
 
-def get_message(prompt, min_new_tokens=16, max_turn=4):
+def get_message(prompt, min_new_tokens=5, max_turn=4):
     prompt = prompt.strip()
     ids = tokenizer(prompt, return_tensors="pt").to(device)
     min_length = ids['input_ids'].shape[1] + min_new_tokens
@@ -28,7 +28,8 @@ def get_message(prompt, min_new_tokens=16, max_turn=4):
         max_new_tokens=128,
         min_length=min_length,
         do_sample=True,
-        top_p=0.7,
+        top_p=0.95,
+        temperature=1.35,
         early_stopping=True
     ) # [0]['generated_text']