ICTNLP

/

LLaMA-Omni2-7B-Bilingual

like 1

Follow

Natural Language Processing Group, Institute of Computing Technology, Chinese Academy of Science 111

How to set up the system prompt for a large language model?
item = self.questions[index] messages = [] speech_list = [] for i, turn in enumerate(item["conversation"]): if i % 2 == 0: messages.append({ "role": "user", "content": DEFAULT_SPEECH_TOKEN, }) speech_list.append(self.load_speech(turn["speech"])) else: messages.append({ "role": "assistant", "content": turn["text"], }) input_ids = self.process_messages(messages) return { "input_ids": input_ids, "speech": speech_list }