E5 base 4k LongEmbed: Extending Embedding Models for Long Context Retrieval. Dawei Zhu, Liang Wang, Nan Yang, Yifan Song, Wenhao Wu, Furu Wei, Sujian Li, arxiv 2024. Github Repo for LongEmbed: https://github.com/dwzhu pku/LongEmbed. This model has 12 layers and the embedding size is 768. Usage Below is an example to encode queries and passages from the MS MARCO passage ranking dataset. Training Details Please refer to our paper at https://arxiv.org/abs/2404.12096.pdf. Note that E5 Base 4k simply expands the position embedding matrix to allow for 4,096 position ids. The embedding vectors for the original pids {0,1,2,...,511} is mapped to represent {0,8,16,...,4088}. Embedding vectors for other pids are trained. So for inputs not exceeding 512 tokens, please multiply the position ids by 8 to maintain the original behavior, as shown in the code above. Benchmark Evaluation Check out unilm/e5 to reproduce evaluation results on the BEIR and MTEB benchmark. Citation If you find our paper or models helpful, please consider cite as follows:
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy