StingNing
StingNing
Could you please provide more details? For example, is the performance fluctuating with respect to the choice of templates? Is your experiment conducted under the zero-shot setting?
> According to the SoftVerbalizer script and my general understanding of what is desired in a frozen PLM training setting, the grouped_parameters_1 of the SoftVerbalizer should be frozen. However, in...
Oh, yes! It is a bug, thanks a lot, we will fix it soon!
Hi, it is an interesting issue, could you please provide more details? For example, did you use multi-GPU techniques such as Data-Parallel when for t5-3b?
Thanks for your contribution，it seems reasonable!
您好，因为根据LLaMA的政策，我们上传的是Delta Weights，需要和原始权重进行合并后使用。
> I observe that the inference script contains an embedded system prompt. Is it also included in training? I wonder how much the performance will be affected if changing the...
您好，感谢关注！我们在这个[issue](https://github.com/thunlp/UltraChat/issues/22)中提供了生成用户的prompt，这里核心的关键是需要将历史对话和prompt进行拼接来获取更好的一致性。我们未来会以Sector 1为例子，整理开放所有的代码，中间结果和元信息，以及prompts。
感谢关注！实际上我们尝试构造了一批中文数据集，但是此数据的质量还没有达到我们的标准，因此选择暂不公开。具体的原因我们会后续查清楚。
> > We instruct the user model with carefully designed prompts to mimic human user behavior > > @ningding97 could you provide these prompts you used? Thanks! Hi, thanks for...