Update app.py

app.py

@@ -12,8 +12,8 @@ def load_model():
     tokenizer = AutoTokenizer.from_pretrained(model_name)
     model = AutoModelForCausalLM.from_pretrained(
         model_name,
-        torch_dtype=torch.float16,
-        device_map="auto"
+        torch_dtype=torch.float32,
+        device_map="cpu"
     )
     return model, tokenizer
 
@@ -79,6 +79,7 @@ def main():
             with st.spinner("Réflexion en cours..."):
                 prompt = format_prompt(question)
                 response = generate_response(prompt, model, tokenizer)
+                response = response.replace('<think>', '')
                 st.markdown(response)
         st.session_state.messages.append({"role": "assistant", "content": response})