update

app.py

@@ -20,6 +20,7 @@ os.environ['HF_HOME'] = '/app/.cache'
 # model = T5ForConditionalGeneration.from_pretrained(model_name)
 # model = AutoModelForCausalLM.from_pretrained("THUDM/longwriter-glm4-9b", trust_remote_code=True)
 # tokenizer = AutoTokenizer.from_pretrained("THUDM/longwriter-glm4-9b", trust_remote_code=True)
+prompt_first ="Résumé en 4 ou 5 phrases de la plainte suivante :"
 
 # Configuration du client OpenAI avec l'API NVIDIA
 client = OpenAI(
@@ -98,8 +99,8 @@ async def generate_text(request: RequestModel):
         model="meta/llama-3.1-8b-instruct", 
         messages=[{"role": "user", "content": instructions_par_defaut + request.text}],
         temperature=0.2,
-        top_p=0.7,
-        max_tokens=1024, 
+        top_p=0.9,
+        # max_tokens=1024, 
         stream=True
     )
     # response = ollama.chat(