Update app.py

app.py

@@ -22,13 +22,8 @@ tokenizer, model = load_model()
 
 
 @spaces.GPU
-def generate(prompt, history):
-    conversation = []
-    for message in history:
-        conversation.append({
-            "role": message["role"],
-            "content": message["content"]
-        })
+def generate(message, history):
+    conversation = history + [message]
     
     inputs = tokenizer.apply_chat_template(
         conversation,
@@ -41,7 +36,7 @@ def generate(prompt, history):
     with torch.no_grad():
         outputs = model.generate(
             **inputs,
-            max_new_tokens=1024,
+            max_new_tokens=512,
         )
     
     generated_tokens = outputs[0, inputs.input_ids.shape[1]:]