FlagEmbedding
FlagEmbedding copied to clipboard
FlagOpen
Retrieval and Retrieval-augmented LLMs
What's the difference between [encode_corpus](https://github.com/FlagOpen/FlagEmbedding/blob/4efa19d7eb6661494280df556cb8e92c9363c4f1/FlagEmbedding/inference/embedder/decoder_only/base.py#L132C9-L132C14) and [encode](https://github.com/FlagOpen/FlagEmbedding/blob/4efa19d7eb6661494280df556cb8e92c9363c4f1/FlagEmbedding/inference/embedder/decoder_only/base.py#L160)?
想请问以下几个问题： 1、encoder类型的模型用的是cls进行相似度分的输出 decoder类型的模型也输出相似度分吗？是怎么输出的？ 2、lightweight的含义是什么？ 3、有各个模型的推理速度数据吗，如果是只输出一个分数，2b的模型应该也不会慢吧
# 请问使用flagEmbedding做向量化的时候，使用bge-m3模型，在高并发下有最佳实践可以参考吗，目前我自己包装了一个使用flagembedding服务来做向量化，在使用多线程的情况下遇到奇奇怪怪的问题，以下是我的代码 ` import os import traceback from concurrent.futures import ThreadPoolExecutor from fastapi import FastAPI, HTTPException from fastapi.responses import JSONResponse from fastapi.encoders import jsonable_encoder import uvicorn import asyncio from log.log_info...
``` E1119 08:26:02.715000 28705 site-packages/torch/distributed/elastic/multiprocessing/api.py:882] failed (exitcode: -11) local_rank: 0 (pid: 28773) of binary: /app/anaconda3/envs/py312/bin/python3.12 Traceback (most recent call last): File "/app/anaconda3/envs/py312/bin/torchrun", line 7, in sys.exit(main()) ^^^^^^ File "/app/anaconda3/envs/py312/lib/python3.12/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line...
hi，我对Activation beacon工作非常感兴趣，想尝试使用hf上的例子跑该模型。但是输出结果是乱码。 使用的例子： messages = [{"role": "user", "content": "Tell me about yourself."}] 输出： Input Length: 24 Output: 'system\nYou are a helpful assistant.\nuser\nTell me about yourself.\nassistant\n - 10.00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000' 我使用的transformer版本是4.45.0，fa版本是2.7.2
Could there be an issue with the parameter settings in my training script? export WANDB_MODE=disabled train_data="\ /home/jovyan/dataws1/bgeft/train_table_data " # set large epochs and small batch size for testing num_train_epochs=1 per_device_train_batch_size=1...
When installing FlagEmbeddings 1.3.2, it will pull the 0.18 version of peft which is incompatible with your library and breaks the model for inference. Can you pin the dependencies so...
Hi, I used `uv add "visual_bge @ git+https://github.com/FlagOpen/FlagEmbedding.git#subdirectory=research/visual_bge"` to install `visual_bge` and I can not import visual_bge. I see: And Should `__init__.py` be under visual_bge/visual_bge?
switched tokenizer.pad to __call__ method for better performance and error avoidance
