FlagEmbedding icon indicating copy to clipboard operation
FlagEmbedding copied to clipboard

Retrieval and Retrieval-augmented LLMs

Results 622 FlagEmbedding issues
Sort by recently updated
recently updated
newest added

# BAAI / bge-reranker-base 模型转为onnx的疑问 如题,当我转换huggingface上提供的模型为onnx时,生成的onnx模型在运行时只输出logits,而不是分类的分数。 ## 转换代码如下 ```python import torch from transformers import BertForSequenceClassification import onnx from transformers import AutoModel import logging logging.basicConfig(level=logging.DEBUG) device = torch.device('cuda:0' if torch.cuda.is_available() else...

Hey there, thank you guys for sharing the model as well. I was curious, are there any plans to support onnx for this model?

Thank you for sharing the M3 model with the community! This is a welcome direction, where you can get multiple representations from the same model! One comment and two questions...

根据技术文档中提到的数据,采用三种方式联合检索的情况下有明显的性能优势。我不太理解这里的混合搜索如何和向量库一起使用。 如果是 RAG 应用,存在大量的文本,是需要提前做嵌入的。这时候是先按 Dense 向量存储,查询的时候先粗筛 TopK * n 然后再对粗筛的数据应用混合搜索筛选出 TopK,再对 TopK 应用 Rerank ? 希望能得到您的解答,非常感谢🙏

作者您好,请问BGE有相关的性能测试报告吗,大概每秒多少token 或 样本 非常感谢您的回答!

下面代码运行很慢,一次向量化需要17秒,不管是CPU还是GPU都很慢,请问这个怎么解决 ``` from multiprocessing import freeze_support from FlagEmbedding import BGEM3FlagModel import time def main(): model = BGEM3FlagModel('BAAI/bge-m3', use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance...

MKQA本身数据集并不包含corpus,请问是如何和nq的corpus做对应的呢

请问有没有类似m3e的api调用方式?

![image](https://github.com/FlagOpen/FlagEmbedding/assets/42883258/08372a16-8d7d-43c7-97fe-6ef818af2ddf) ![image](https://github.com/FlagOpen/FlagEmbedding/assets/42883258/4410ba82-98bf-4399-bafe-be9a2afc05ec) 如上图所示,这个结果代表什么意思?

您的模型是按照语言分为了bge-large-en、bge-large-zh,但我测试了bge-large-zh也能很好地对英文计算相似度。想确认一下bge-large-zh是否支持中英文双语?如果支持,bge-large-zh在英文数据集上测试的效果如何?