silvercherry

这个训练后要怎么测试呢

Have you compared your results with those of the paper

1

I think this is a very good job, I don’t know what your results are, is it similar to the results in the paper?

llama3-8b扩充词表训练RuntimeError: CUDA error: device-side assert triggered

1

我扩充了llama3的训练词表，并修改config中tokenizer和model中响应部分，但是报了RuntimeError: CUDA error: device-side assert triggered这样的错误。具体错误见图 tokenizer = dict( type=AutoTokenizer.from_pretrained, pretrained_model_name_or_path=new_tokens, trust_remote_code=True, padding_side='right') model = dict( type=SupervisedFinetune, use_varlen_attn=use_varlen_attn, llm=dict( type=AutoModelForCausalLM.from_pretrained, pretrained_model_name_or_path=pretrained_model_name_or_path, trust_remote_code=True, torch_dtype=torch.float16), tokenizer=tokenizer) 并且在看log的时候，模型的embedding也做了resize（Resized token embeddings from 128256 to...

what is the different between target and universal？

2

I checked the code and found nothing different during training，How should I do target attack?