BAAI/bge-reranker-v2.5-gemma2-lightweight

git clone https://github.com/FlagOpen/FlagEmbedding.git
cd FlagEmbedding
pip install -e .

from FlagEmbedding import LightWeightFlagLLMReranker
reranker = LightWeightFlagLLMReranker('BAAI/bge-reranker-v2.5-gemma2-lightweight', use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation

score = reranker.compute_score(['query', 'passage'], cutoff_layers=[28], compress_ratio=2, compress_layer=[24, 40]) # 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], compress_ratio=2, compress_layer=[24, 40])
print(scores)

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

def last_logit_pool(logits: torch.Tensor,
                    attention_mask: torch.Tensor) -> torch.Tensor:
    left_padding = (attention_mask[:, -1].sum() == attention_mask.shape[0])
    if left_padding:
        return logits[:, -1]
    else:
        sequence_lengths = attention_mask.sum(dim=1) - 1
        batch_size = logits.shape[0]
        return torch.stack([logits[i, sequence_lengths[i]] for i in range(batch_size)], dim=0)

def get_inputs(pairs, tokenizer, prompt=None, max_length=1024):
    if prompt is None:
        prompt = "Predict whether passage B contains an answer to query A."
    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 = []
    query_lengths = []
    prompt_lengths = []
    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)
        query_lengths.append(len([tokenizer.bos_token_id] + query_inputs['input_ids'] + sep_inputs))
        prompt_lengths.append(len(sep_inputs + prompt_inputs))
        
    return tokenizer.pad(
            inputs,
            padding=True,
            max_length=max_length + len(sep_inputs) + len(prompt_inputs),
            pad_to_multiple_of=8,
            return_tensors='pt',
    ), query_lengths, prompt_lengths

tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-reranker-v2.5-gemma2-lightweight', trust_remote_code=True)
tokenizer.padding_side = 'right'
model = AutoModelForCausalLM.from_pretrained('BAAI/bge-reranker-v2.5-gemma2-lightweight', trust_remote_code=True)
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, query_lengths, prompt_lengths = get_inputs(pairs, tokenizer)
    inputs = inputs.to(model.device)
    outputs = model(**inputs,
                    return_dict=True,
                    cutoff_layers=[28],
                    compress_ratio=2,
                    compress_layer=[24, 40],
                    query_lengths=query_lengths,
                    prompt_lengths=prompt_lengths)
    scores = []
    for i in range(len(outputs.logits)):
        logits = last_logit_pool(outputs.logits[i], outputs.attention_masks[i])
        scores.append(logits.cpu().float().tolist())
    print(scores)

"auto_map": {
    "AutoConfig": "gemma_config.CostWiseGemmaConfig",
    "AutoModel": "gemma_model.CostWiseGemmaModel",
    "AutoModelForCausalLM": "gemma_model.CostWiseGemmaForCausalLM"
  },

@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}
}