FlagEmbedding
FlagEmbedding copied to clipboard
FlagOpen
Retrieval and Retrieval-augmented LLMs
使用的bge-m3 embedding模型，想问下跑完MultiVector and All Rerank这个评估过程以后，embedding+rerank的召回分数还降低了这个正常嘛，两个都是原本的比原embedding分数低，两个都是微调过的比微调过的embedding分数低
Is there any code that evaluates the miracl dataset with the Ndcg@10 metric? Or, I know that in order to evaluate Ndcg on the miracl dataset, the similarity between positives...
Is there any BGE model that can be used for multi-label text classification (predefined labels) by adding some dense layers for classification (similar to BERT base uncased) instead of Sentence...
微调问题
(ft_emb) b405@b405-CVN-Z790-GAMING-FROZEN:/media/b405/新加卷1/Workspace_linux/b405/ZH/llmProjects/FlagEmbedding$ torchrun --nproc_per_node 1 \ -m FlagEmbedding.reranker.run \ --output_dir /media/b405/新加卷1/Workspace_linux/b405/ZH/llmProjects/FlagEmbedding/fine_tune_rerank_models/beg_rerank_ft \ --model_name_or_path /media/b405/新加卷1/Workspace_linux/b405/ZH/embeddingModels/bge-reranker-large/ \ --train_data /media/b405/新加卷1/Workspace_linux/b405/ZH/llmProjects/FlagEmbedding/post_process_embedding_finetune_dataset.jsonl \ --learning_rate 6e-5 \ --fp16 \ --num_train_epochs 5 \ --per_device_train_batch_size 10 \ --gradient_accumulation_steps...
Refer to the following method to use the m3 model, an error will occur when this service is called in parallel. ``` from FlagEmbedding import BGEM3FlagModel model = BGEM3FlagModel('BAAI/bge-m3', use_fp16=True)...
你好，请教一下我想用自己的tokenizer【chatglm的tokenizer】来训练bge模型应该怎么做呢？我现在的想法是把【chatglm的tokenizer】改成类似bert-tokenizer，这样可行吗？
from FlagEmbedding import FlagReranker reranker = FlagReranker('/bge-reranker-v2-m3/bge-reranker-v2-m3', use_fp16=True) score = reranker.compute_score(['Jc Sweet House', 'Candy House'], normalize=True) print(score)#0.14354750671079064 score = reranker.compute_score(['jc sweet house', 'candy house'], normalize=True) print(score)#0.8033281311678119
Can BGE rerank be finetuned in an instruction style, for example: "tool": { "query": "Transform this user request for fetching helpful tool descriptions: ", "key": "Transform this tool description for...