jackwwy

Hello, I would like to ask, when you are training the model, do you only use the first round of dialogue from the ultrachat_200k?

1

`def load_and_process_data_ultrachat(dataset_name, split): try: dataset = load_dataset(dataset_name, split=split) reformatted_data = [{ 'generated': [message['messages'][0], {"role": "assistant", "content": ""}], 'real': [message['messages'][0], message['messages'][1]] } for message in dataset] return reformatted_data except Exception as...

你好 请问如果我要复现主实验的结果，需要我对100个遗忘目标中的每一个遗忘目标进行训练，评测。然后取平均值吗？如果这样做，我觉得太消耗时间了。另外，对于GA算法，你是使用llama-factory的pre-train训练机制吗？因为我看到你的dataset_info.json文件中的格式是 "columns": { "prompt": "text" }