--- license: apache-2.0 pipeline_tag: text-classification tags: - transformers - sentence-transformers - text-embeddings-inference language: - multilingual --- # Reranker **More details please refer to our Github: [FlagEmbedding](https://github.com/FlagOpen/FlagEmbedding/tree/master).** - [Model List](#model-list) - [Usage](#usage) - [Fine-tuning](#fine-tune) - [Evaluation](#evaluation) - [Citation](#citation) Different from embedding model, reranker uses question and document as input and directly output similarity instead of embedding. You can get a relevance score by inputting query and passage to the reranker. And the score can be mapped to a float value in [0,1] by sigmoid function. ## Model List | Model | Base model | Language | layerwise | feature | |:--------------------------------------------------------------------------|:--------:|:-----------------------------------------------------------------------------------------------------------------------------------:|:----------------------------------------------------------------------------------------------:|:----------------------------------------------------------------------------------------------:| | [BAAI/bge-reranker-base](https://huggingface.co./BAAI/bge-reranker-base) | [xlm-roberta-base](https://huggingface.co./xlm-roberta-base) | Chinese and English | - | Lightweight reranker model, easy to deploy, with fast inference. | | [BAAI/bge-reranker-large](https://huggingface.co./BAAI/bge-reranker-large) | [xlm-roberta-large](https://huggingface.co./FacebookAI/xlm-roberta-large) | Chinese and English | - | Lightweight reranker model, easy to deploy, with fast inference. | | [BAAI/bge-reranker-v2-m3](https://huggingface.co./BAAI/bge-reranker-v2-m3) | [bge-m3](https://huggingface.co./BAAI/bge-m3) | Multilingual | - | Lightweight reranker model, possesses strong multilingual capabilities, easy to deploy, with fast inference. | | [BAAI/bge-reranker-v2-gemma](https://huggingface.co./BAAI/bge-reranker-v2-gemma) | [gemma-2b](https://huggingface.co./google/gemma-2b) | Multilingual | - | Suitable for multilingual contexts, performs well in both English proficiency and multilingual capabilities. | | [BAAI/bge-reranker-v2-minicpm-layerwise](https://huggingface.co./BAAI/bge-reranker-v2-minicpm-layerwise) | [MiniCPM-2B-dpo-bf16](https://huggingface.co./openbmb/MiniCPM-2B-dpo-bf16) | Multilingual | 8-40 | Suitable for multilingual contexts, performs well in both English and Chinese proficiency, allows freedom to select layers for output, facilitating accelerated inference. | You can select the model according your senario and resource. - For **multilingual**, utilize [BAAI/bge-reranker-v2-m3](https://huggingface.co./BAAI/bge-reranker-v2-m3) and [BAAI/bge-reranker-v2-gemma](https://huggingface.co./BAAI/bge-reranker-v2-gemma) - For **Chinese or English**, utilize [BAAI/bge-reranker-v2-m3](https://huggingface.co./BAAI/bge-reranker-v2-m3) and [BAAI/bge-reranker-v2-minicpm-layerwise](https://huggingface.co./BAAI/bge-reranker-v2-minicpm-layerwise). - For **efficiency**, utilize [BAAI/bge-reranker-v2-m3](https://huggingface.co./BAAI/bge-reranker-v2-m3) and the low layer of [BAAI/bge-reranker-v2-minicpm-layerwise](https://huggingface.co./BAAI/bge-reranker-v2-minicpm-layerwise). - For better performance, recommand [BAAI/bge-reranker-v2-minicpm-layerwise](https://huggingface.co./BAAI/bge-reranker-v2-minicpm-layerwise) and [BAAI/bge-reranker-v2-gemma](https://huggingface.co./BAAI/bge-reranker-v2-gemma) ## Usage ### Using FlagEmbedding ``` pip install -U FlagEmbedding ``` #### For normal reranker (bge-reranker-base / bge-reranker-large / bge-reranker-v2-m3 ) Get relevance scores (higher scores indicate more relevance): ```python from FlagEmbedding import FlagReranker reranker = FlagReranker('BAAI/bge-reranker-v2-m3', use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation score = reranker.compute_score(['query', 'passage']) print(score) # -5.65234375 # You can map the scores into 0-1 by set "normalize=True", which will apply sigmoid function to the score score = reranker.compute_score(['query', 'passage'], normalize=True) print(score) # 0.003497010252573502 scores = reranker.compute_score([['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']]) print(scores) # [-8.1875, 5.26171875] # You can map the scores into 0-1 by set "normalize=True", which will apply sigmoid function to the score scores = reranker.compute_score([['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']], normalize=True) print(scores) # [0.00027803096387751553, 0.9948403768236574] ``` #### For LLM-based reranker ```python from FlagEmbedding import FlagLLMReranker reranker = FlagLLMReranker('BAAI/bge-reranker-v2-gemma', use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation # reranker = FlagLLMReranker('BAAI/bge-reranker-v2-gemma', use_bf16=True) # You can also set use_bf16=True to speed up computation with a slight performance degradation score = reranker.compute_score(['query', 'passage']) print(score) scores = reranker.compute_score([['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']]) print(scores) ``` #### For LLM-based layerwise reranker ```python from FlagEmbedding import LayerWiseFlagLLMReranker reranker = LayerWiseFlagLLMReranker('BAAI/bge-reranker-v2-minicpm-layerwise', use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation # reranker = LayerWiseFlagLLMReranker('BAAI/bge-reranker-v2-minicpm-layerwise', use_bf16=True) # You can also set use_bf16=True to speed up computation with a slight performance degradation score = reranker.compute_score(['query', 'passage'], cutoff_layers=[28]) # Adjusting 'cutoff_layers' to pick which layers are used for computing the score. print(score) scores = reranker.compute_score([['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']], cutoff_layers=[28]) print(scores) ``` ### Using Huggingface transformers #### For normal reranker (bge-reranker-base / bge-reranker-large / bge-reranker-v2-m3 ) Get relevance scores (higher scores indicate more relevance): ```python import torch from transformers import AutoModelForSequenceClassification, AutoTokenizer tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-reranker-v2-m3') model = AutoModelForSequenceClassification.from_pretrained('BAAI/bge-reranker-v2-m3') model.eval() pairs = [['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']] with torch.no_grad(): inputs = tokenizer(pairs, padding=True, truncation=True, return_tensors='pt', max_length=512) scores = model(**inputs, return_dict=True).logits.view(-1, ).float() print(scores) ``` #### For LLM-based reranker ```python import torch from transformers import AutoModelForCausalLM, AutoTokenizer def get_inputs(pairs, tokenizer, prompt=None, max_length=1024): if prompt is None: prompt = "Given a query A and a passage B, determine whether the passage contains an answer to the query by providing a prediction of either 'Yes' or 'No'." sep = "\n" prompt_inputs = tokenizer(prompt, return_tensors=None, add_special_tokens=False)['input_ids'] sep_inputs = tokenizer(sep, return_tensors=None, add_special_tokens=False)['input_ids'] inputs = [] for query, passage in pairs: query_inputs = tokenizer(f'A: {query}', return_tensors=None, add_special_tokens=False, max_length=max_length * 3 // 4, truncation=True) passage_inputs = tokenizer(f'B: {passage}', return_tensors=None, add_special_tokens=False, max_length=max_length, truncation=True) item = tokenizer.prepare_for_model( [tokenizer.bos_token_id] + query_inputs['input_ids'], sep_inputs + passage_inputs['input_ids'], truncation='only_second', max_length=max_length, padding=False, return_attention_mask=False, return_token_type_ids=False, add_special_tokens=False ) item['input_ids'] = item['input_ids'] + sep_inputs + prompt_inputs item['attention_mask'] = [1] * len(item['input_ids']) inputs.append(item) return tokenizer.pad( inputs, padding=True, max_length=max_length + len(sep_inputs) + len(prompt_inputs), pad_to_multiple_of=8, return_tensors='pt', ) tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-reranker-v2-gemma') model = AutoModelForCausalLM.from_pretrained('BAAI/bge-reranker-v2-gemma') yes_loc = tokenizer('Yes', add_special_tokens=False)['input_ids'][0] model.eval() pairs = [['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']] with torch.no_grad(): inputs = get_inputs(pairs, tokenizer) scores = model(**inputs, return_dict=True).logits[:, -1, yes_loc].view(-1, ).float() print(scores) ``` #### For LLM-based layerwise reranker ```python import torch from transformers import AutoModelForCausalLM, AutoTokenizer def get_inputs(pairs, tokenizer, prompt=None, max_length=1024): if prompt is None: prompt = "Given a query A and a passage B, determine whether the passage contains an answer to the query by providing a prediction of either 'Yes' or 'No'." sep = "\n" prompt_inputs = tokenizer(prompt, return_tensors=None, add_special_tokens=False)['input_ids'] sep_inputs = tokenizer(sep, return_tensors=None, add_special_tokens=False)['input_ids'] inputs = [] for query, passage in pairs: query_inputs = tokenizer(f'A: {query}', return_tensors=None, add_special_tokens=False, max_length=max_length * 3 // 4, truncation=True) passage_inputs = tokenizer(f'B: {passage}', return_tensors=None, add_special_tokens=False, max_length=max_length, truncation=True) item = tokenizer.prepare_for_model( [tokenizer.bos_token_id] + query_inputs['input_ids'], sep_inputs + passage_inputs['input_ids'], truncation='only_second', max_length=max_length, padding=False, return_attention_mask=False, return_token_type_ids=False, add_special_tokens=False ) item['input_ids'] = item['input_ids'] + sep_inputs + prompt_inputs item['attention_mask'] = [1] * len(item['input_ids']) inputs.append(item) return tokenizer.pad( inputs, padding=True, max_length=max_length + len(sep_inputs) + len(prompt_inputs), pad_to_multiple_of=8, return_tensors='pt', ) tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-reranker-v2-minicpm-layerwise', trust_remote_code=True) model = AutoModelForCausalLM.from_pretrained('BAAI/bge-reranker-v2-minicpm-layerwise', trust_remote_code=True, torch_dtype=torch.bfloat16) model = model.to('cuda') model.eval() pairs = [['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']] with torch.no_grad(): inputs = get_inputs(pairs, tokenizer).to(model.device) all_scores = model(**inputs, return_dict=True, cutoff_layers=[28]) all_scores = [scores[:, -1].view(-1, ).float() for scores in all_scores[0]] print(all_scores) ``` ## Fine-tune ### Data Format Train data should be a json file, where each line is a dict like this: ``` {"query": str, "pos": List[str], "neg":List[str], "prompt": str} ``` `query` is the query, and `pos` is a list of positive texts, `neg` is a list of negative texts, `prompt` indicates the relationship between query and texts. If you have no negative texts for a query, you can random sample some from the entire corpus as the negatives. See [toy_finetune_data.jsonl](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/llm_reranker/toy_finetune_data.jsonl) for a toy data file. ### Train You can fine-tune the reranker with the following code: **For llm-based reranker** ```shell torchrun --nproc_per_node {number of gpus} \ -m FlagEmbedding.llm_reranker.finetune_for_instruction.run \ --output_dir {path to save model} \ --model_name_or_path google/gemma-2b \ --train_data ./toy_finetune_data.jsonl \ --learning_rate 2e-4 \ --num_train_epochs 1 \ --per_device_train_batch_size 1 \ --gradient_accumulation_steps 16 \ --dataloader_drop_last True \ --query_max_len 512 \ --passage_max_len 512 \ --train_group_size 16 \ --logging_steps 1 \ --save_steps 2000 \ --save_total_limit 50 \ --ddp_find_unused_parameters False \ --gradient_checkpointing \ --deepspeed stage1.json \ --warmup_ratio 0.1 \ --bf16 \ --use_lora True \ --lora_rank 32 \ --lora_alpha 64 \ --use_flash_attn True \ --target_modules q_proj k_proj v_proj o_proj ``` **For llm-based layerwise reranker** ```shell torchrun --nproc_per_node {number of gpus} \ -m FlagEmbedding.llm_reranker.finetune_for_layerwise.run \ --output_dir {path to save model} \ --model_name_or_path openbmb/MiniCPM-2B-dpo-bf16 \ --train_data ./toy_finetune_data.jsonl \ --learning_rate 2e-4 \ --num_train_epochs 1 \ --per_device_train_batch_size 1 \ --gradient_accumulation_steps 16 \ --dataloader_drop_last True \ --query_max_len 512 \ --passage_max_len 512 \ --train_group_size 16 \ --logging_steps 1 \ --save_steps 2000 \ --save_total_limit 50 \ --ddp_find_unused_parameters False \ --gradient_checkpointing \ --deepspeed stage1.json \ --warmup_ratio 0.1 \ --bf16 \ --use_lora True \ --lora_rank 32 \ --lora_alpha 64 \ --use_flash_attn True \ --target_modules q_proj k_proj v_proj o_proj \ --start_layer 8 \ --head_multi True \ --head_type simple \ --lora_extra_parameters linear_head ``` Our rerankers are initialized from [google/gemma-2b](https://huggingface.co./google/gemma-2b) (for llm-based reranker) and [openbmb/MiniCPM-2B-dpo-bf16](https://huggingface.co./openbmb/MiniCPM-2B-dpo-bf16) (for llm-based layerwise reranker), and we train it on a mixture of multilingual datasets: - [bge-m3-data](https://huggingface.co./datasets/Shitao/bge-m3-data) - [quora train data](https://huggingface.co./datasets/quora) - [fever train data](https://fever.ai/dataset/fever.html) ## Evaluation - llama-index. ![image-20240317193909373](./assets/llama-index.png) - BEIR. rereank the top 100 results from bge-en-v1.5 large. ![image-20240317174633333](./assets/BEIR-bge-en-v1.5.png) rereank the top 100 results from e5 mistral 7b instruct. ![image-20240317172949713](./assets/BEIR-e5-mistral.png) - CMTEB-retrieval. It rereank the top 100 results from bge-zh-v1.5 large. ![image-20240317173026235](./assets/CMTEB-retrieval-bge-zh-v1.5.png) - miracl (multi-language). It rereank the top 100 results from bge-m3. ![image-20240317173117639](./assets/miracl-bge-m3.png) ## Citation If you find this repository useful, please consider giving a star and citation ```bibtex @misc{li2023making, title={Making Large Language Models A Better Foundation For Dense Retrieval}, author={Chaofan Li and Zheng Liu and Shitao Xiao and Yingxia Shao}, year={2023}, eprint={2312.15503}, archivePrefix={arXiv}, primaryClass={cs.CL} } @misc{chen2024bge, title={BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation}, author={Jianlv Chen and Shitao Xiao and Peitian Zhang and Kun Luo and Defu Lian and Zheng Liu}, year={2024}, eprint={2402.03216}, archivePrefix={arXiv}, primaryClass={cs.CL} } ```