Update app.py

app.py

@@ -60,7 +60,7 @@ SYSTEM_PROMPT = "You are a friendly Chatbot."
 @spaces.GPU
 def generate(prompt, history):
     messages = [
-        {"role": "system", "content": SYSTEM_PROMPT},
+        # {"role": "system", "content": SYSTEM_PROMPT},
         {"role": "user", "content": prompt}
     ]
     # text = tokenizer.apply_chat_template(
@@ -87,10 +87,10 @@ def generate(prompt, history):
         # tokenize=True,
         return_dict=True,
         return_tensors="pt",
-    ).to(model.device)
+    )
     outputs = model.generate(
-        **inputs,
-        max_new_tokens=512,
+        **inputs.to(model.device),
+        max_new_tokens=100,
     )
     response = processor.batch_decode(outputs[:, inputs["input_ids"].shape[-1]:])[0]