Yifan Zhu
Results
2
comments of
Yifan Zhu
Very helpful discussion. I successfully ran the inference on 4*RTX 4090 by following your modifications. Thx for you all : )
流程是差不多的。简单去理解的话，前面还是搜索引擎这套链路，最后把排序靠前的 top k 文档丢给 LLM 作为 context，让它结合 context 去回答用户的 query