update

app.py

@@ -33,6 +33,8 @@ def predict(comment):
 	outputs = model.generate(input_ids=inputs["input_ids"].to("cuda"), max_new_tokens=400, pad_token_id=tokenizer.eos_token_id)
 	outputs = tokenizer.batch_decode(outputs)[0]
 	outputs = outputs.split("[/INST]")[1]
+	outputs = outputs.split("[INST]")[0]
+	outputs = outputs.split("</s>")[0]
 	return outputs