Update app.py

app.py

@@ -47,7 +47,8 @@ def load_models():
     classification_model.eval()
     
     # 2. Load text generation model
-    llm_model_name = "Qwen/Qwen3-0.6B"
+    # llm_model_name = "Qwen/Qwen3-0.6B"
+    llm_model_name = "meta-llama/Meta-Llama-3.1-8B-Instruct"
     tokenizer = AutoTokenizer.from_pretrained(llm_model_name)
     llm_model = AutoModelForCausalLM.from_pretrained(
         llm_model_name,
@@ -143,7 +144,7 @@ def generate_response(message, chat_history, analysis_results):
         model_inputs = tokenizer([text], return_tensors="pt").to(llm_model.device)
         generated_ids = llm_model.generate(
             **model_inputs,
-            max_new_tokens=300,
+            max_new_tokens=256,
             do_sample=True,
             temperature=0.7,
         )