Keming (Luke) Lu

icon company Qwen icon location Beijing icon location - Alignment Researcher & LLM Developer at Qwen, Alibaba Inc. - MS at USC, Information Extraction - BEng at Tsinghua University, Automation & Statistics

Results 7 comments of Keming (Luke) Lu

Will you release data and corresponding tags?

Thanks for your interest. We will work on the release process of 6k filtered data and open-sourced annotated data as much as possible. Stay tune guys

How should SamplingParams be configured?

Thanks for your comment! The SamplingParams for Qwen-1.8B tagger is greedy (temperature=0) and max_tokens can be set to 512 or larger.

"This repository is over its data quota"，无法下载git lfs的内容

Thanks for raising this issue. This error is due to the emerging download of this repo. We are working on pushing our data to HuggingFace and ModelScope.

"This repository is over its data quota"，无法下载git lfs的内容

Sorry for late reply. Please kindly download the data from this shared link http://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/lukeming.lkm/public/ditto_data.zip

麻烦数据提供下，gitlfs没法下载，感谢

Thanks for your kind remind. Please kindly download data from this shared link http://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/lukeming.lkm/public/ditto_data.zip

[character_AI_open] Related works with open-source code, data, models, and it's two months ahead.

Looks good. Cant wait for your latest progress.

Details about Qwen-1_8B Instagger.

Hi litsh, the tagger is trained with vanilla prompts and a default system prompt "You are a helpful assistant.", so no other template prompts are used for inference. For example,...