Skip to content

Reranking

rankify.models.reranking

HF_PRE_DEFIND_MODELS = {'upr': {'t5-small': 'google/t5-small-lm-adapt', 't5-base': 'google/t5-base-lm-adapt', 't5-large': 'google/t5-large-lm-adapt', 't0-3b': 'bigscience/T0_3B', 't0-11b': 'bigscience/T0', 'gpt-neo-2.7b': 'EleutherAI/gpt-neo-2.7B', 'gpt-j-6b': 'EleutherAI/gpt-j-6b', 'gpt2': 'openai-community/gpt2', 'gpt2-medium': 'openai-community/gpt2-medium', 'gpt2-large': 'openai-community/gpt2-large', 'gpt2-xl': 'openai-community/gpt2-xl', 'flan-t5-xl': 'google/flan-t5-xl'}, 'rankgpt-api': {'gpt-3.5': 'gpt-3.5', 'gpt-4': 'gpt-4o', 'gpt-4-mini': 'gpt-4o-mini', 'llamav3.1-8b': 'llamav3.1-8b', 'llamav3.1-70b': 'llamav3.1-70b', 'claude-3-5': 'claude-3-5'}, 'rankgpt': {'llamav3.1-8b': 'meta-llama/Meta-Llama-3.1-8B-Instruct', 'llamav3.1-70b': 'meta-llama/Llama-3.1-70B-Instruct', 'Llama-3.2-1B': 'meta-llama/Llama-3.2-1B-Instruct', 'Llama-3.2-3B': 'meta-llama/Llama-3.2-3B-Instruct', 'Qwen2.5-7B': 'Qwen/Qwen2.5-7B', 'Mistral-7B-Instruct-v0.2': 'mistralai/Mistral-7B-Instruct-v0.2', 'Mistral-7B-Instruct-v0.3': 'mistralai/Mistral-7B-Instruct-v0.3'}, 'flashrank': {'ms-marco-TinyBERT-L-2-v2': 'ms-marco-TinyBERT-L-2-v2', 'ms-marco-MiniLM-L-12-v2': 'ms-marco-MiniLM-L-12-v2', 'ms-marco-MultiBERT-L-12': 'ms-marco-MultiBERT-L-12', 'rank-T5-flan': 'rank-T5-flan', 'ce-esci-MiniLM-L12-v2': 'ce-esci-MiniLM-L12-v2', 'rank_zephyr_7b_v1_full': 'rank_zephyr_7b_v1_full', 'miniReranker_arabic_v1': 'miniReranker_arabic_v1'}, 'flashrank-model-file': {'ms-marco-TinyBERT-L-2-v2': 'flashrank-TinyBERT-L-2-v2.onnx', 'ms-marco-MiniLM-L-12-v2': 'flashrank-MiniLM-L-12-v2_Q.onnx', 'ms-marco-MultiBERT-L-12': 'flashrank-MultiBERT-L12_Q.onnx', 'rank-T5-flan': 'flashrank-rankt5_Q.onnx', 'ce-esci-MiniLM-L12-v2': 'flashrank-ce-esci-MiniLM-L12-v2_Q.onnx', 'rank_zephyr_7b_v1_full': 'rank_zephyr_7b_v1_full.Q4_K_M.gguf', 'miniReranker_arabic_v1': 'miniReranker_arabic_v1.onnx'}, 'monot5': {'monot5-base-msmarco': 'castorini/monot5-base-msmarco', 'monot5-base-msmarco-10k': 'castorini/monot5-base-msmarco-10k', 'monot5-large-msmarco': 'castorini/monot5-large-msmarco', 'monot5-large-msmarco-10k': 'castorini/monot5-large-msmarco-10k', 'monot5-base-med-msmarco': 'castorini/monot5-base-med-msmarco', 'monot5-3b-med-msmarco': 'castorini/monot5-3b-med-msmarco', 'monot5-3b-msmarco-10k': 'castorini/monot5-3b-msmarco-10k', 'mt5-base-en-msmarco': 'unicamp-dl/mt5-base-en-msmarco', 'ptt5-base-pt-msmarco-10k-v2': 'unicamp-dl/ptt5-base-pt-msmarco-10k-v2', 'ptt5-base-pt-msmarco-100k-v2': 'unicamp-dl/ptt5-base-pt-msmarco-100k-v2', 'ptt5-base-en-pt-msmarco-100k-v2': 'unicamp-dl/ptt5-base-en-pt-msmarco-100k-v2', 'mt5-base-en-pt-msmarco-v2': 'unicamp-dl/mt5-base-en-pt-msmarco-v2', 'mt5-base-mmarco-v2': 'unicamp-dl/mt5-base-mmarco-v2', 'mt5-base-en-pt-msmarco-v1': 'unicamp-dl/mt5-base-en-pt-msmarco-v1', 'mt5-base-mmarco-v1': 'unicamp-dl/mt5-base-mmarco-v1', 'ptt5-base-pt-msmarco-10k-v1': 'unicamp-dl/ptt5-base-pt-msmarco-10k-v1', 'ptt5-base-pt-msmarco-100k-v1': 'unicamp-dl/ptt5-base-pt-msmarco-100k-v1', 'ptt5-base-en-pt-msmarco-10k-v1': 'unicamp-dl/ptt5-base-en-pt-msmarco-10k-v1', 'mt5-3B-mmarco-en-pt': 'unicamp-dl/mt5-3B-mmarco-en-pt', 'mt5-13b-mmarco-100k': 'unicamp-dl/mt5-13b-mmarco-100k', 'monoptt5-small': 'unicamp-dl/monoptt5-small', 'monoptt5-base': 'unicamp-dl/monoptt5-base', 'monoptt5-large': 'unicamp-dl/monoptt5-large', 'monoptt5-3b': 'unicamp-dl/monoptt5-3b'}, 'rankt5': {'rankt5-base': 'Soyoung97/RankT5-base', 'rankt5-large': 'Soyoung97/RankT5-large', 'rankt5-3b': 'Soyoung97/RankT5-3b'}, 'listt5': {'listt5-base': 'Soyoung97/ListT5-base', 'listt5-3b': 'Soyoung97/ListT5-3b'}, 'inranker': {'inranker-small': 'unicamp-dl/InRanker-small', 'inranker-base': 'unicamp-dl/InRanker-base', 'inranker-3b': 'unicamp-dl/InRanker-3B'}, 'apiranker': {'cohere': 'cohere', 'jina': 'jina', 'voyage': 'voyage', 'mixedbread.ai': 'mixedbread.ai'}, 'transformer_ranker': {'mxbai-rerank-xsmall': 'mixedbread-ai/mxbai-rerank-xsmall-v1', 'mxbai-rerank-base': 'mixedbread-ai/mxbai-rerank-base-v1', 'mxbai-rerank-large': 'mixedbread-ai/mxbai-rerank-large-v1', 'bge-reranker-base': 'BAAI/bge-reranker-base', 'bge-reranker-large': 'BAAI/bge-reranker-large', 'bge-reranker-v2-m3': 'BAAI/bge-reranker-v2-m3', 'bce-reranker-base': 'maidalun1020/bce-reranker-base_v1', 'jina-reranker-tiny': 'jinaai/jina-reranker-v1-tiny-en', 'jina-reranker-turbo': 'jinaai/jina-reranker-v1-turbo-en', 'jina-reranker-base-multilingual': 'jinaai/jina-reranker-v2-base-multilingual', 'gte-multilingual-reranker-base': 'Alibaba-NLP/gte-multilingual-reranker-base', 'camembert-base-mmarcoFR': 'antoinelouis/crossencoder-camembert-base-mmarcoFR', 'camembert-large-mmarcoFR': 'antoinelouis/crossencoder-camembert-large-mmarcoFR', 'camemberta-base-mmarcoFR': 'antoinelouis/crossencoder-camemberta-base-mmarcoFR', 'distilcamembert-mmarcoFR': 'antoinelouis/crossencoder-distilcamembert-mmarcoFR', 'cross-encoder-mmarco-mMiniLMv2-L12-H384-v1': 'corrius/cross-encoder-mmarco-mMiniLMv2-L12-H384-v1', 'nli-deberta-v3-large': 'cross-encoder/nli-deberta-v3-large', 'ms-marco-MiniLM-L-12-v2': 'cross-encoder/ms-marco-MiniLM-L-12-v2', 'ms-marco-MiniLM-L-6-v2': 'cross-encoder/ms-marco-MiniLM-L-6-v2', 'ms-marco-MiniLM-L-4-v2': 'cross-encoder/ms-marco-MiniLM-L-4-v2', 'ms-marco-MiniLM-L-2-v2': 'cross-encoder/ms-marco-MiniLM-L-2-v2', 'ms-marco-TinyBERT-L-2-v2': 'cross-encoder/ms-marco-TinyBERT-L-2-v2', 'ms-marco-electra-base': 'cross-encoder/ms-marco-electra-base', 'ms-marco-TinyBERT-L-6': 'cross-encoder/ms-marco-TinyBERT-L-6', 'ms-marco-TinyBERT-L-4': 'cross-encoder/ms-marco-TinyBERT-L-4', 'ms-marco-TinyBERT-L-2': 'cross-encoder/ms-marco-TinyBERT-L-2', 'msmarco-MiniLM-L12-en-de-v1': 'cross-encoder/msmarco-MiniLM-L12-en-de-v1', 'msmarco-MiniLM-L6-en-de-v1': 'cross-encoder/msmarco-MiniLM-L6-en-de-v1'}, 'llm_layerwise_ranker': {'bge-multilingual-gemma2': 'BAAI/bge-multilingual-gemma2', 'bge-reranker-v2-gemma': 'BAAI/bge-reranker-v2-gemma', 'bge-reranker-v2-minicpm-layerwise': 'BAAI/bge-reranker-v2-minicpm-layerwise', 'bge-reranker-v2.5-gemma2-lightweight': 'BAAI/bge-reranker-v2.5-gemma2-lightweight'}, 'first_ranker': {'First-Model': 'rryisthebest/First_Model', 'Llama-3-8B': 'meta-llama/Meta-Llama-3-8B-Instruct'}, 'lit5dist': {'LiT5-Distill-base': 'castorini/LiT5-Distill-base', 'LiT5-Distill-large': 'castorini/LiT5-Distill-large', 'LiT5-Distill-xl': 'castorini/LiT5-Distill-xl', 'LiT5-Distill-base-v2': 'castorini/LiT5-Distill-base-v2', 'LiT5-Distill-large-v2': 'castorini/LiT5-Distill-large-v2', 'LiT5-Distill-xl-v2': 'castorini/LiT5-Distill-xl-v2'}, 'lit5score': {'LiT5-Score-base': 'castorini/LiT5-Score-base', 'LiT5-Score-large': 'castorini/LiT5-Score-large', 'LiT5-Score-xl': 'castorini/LiT5-Score-xl'}, 'vicuna_reranker': {'rank_vicuna_7b_v1': 'castorini/rank_vicuna_7b_v1', 'rank_vicuna_7b_v1_noda': 'castorini/rank_vicuna_7b_v1_noda', 'rank_vicuna_7b_v1_fp16': 'castorini/rank_vicuna_7b_v1_fp16', 'rank_vicuna_7b_v1_noda_fp16': 'castorini/rank_vicuna_7b_v1_noda_fp16'}, 'zephyr_reranker': {'rank_zephyr_7b_v1_full': 'castorini/rank_zephyr_7b_v1_full'}, 'blender_reranker': {'PairRM': 'llm-blender/PairRM'}, 'splade_reranker': {'splade-cocondenser': 'naver/splade-cocondenser-ensembledistil'}, 'sentence_transformer_reranker': {'all-MiniLM-L6-v2': 'all-MiniLM-L6-v2', 'gtr-t5-base': 'sentence-transformers/gtr-t5-base', 'gtr-t5-large': 'sentence-transformers/gtr-t5-large', 'gtr-t5-xl': 'sentence-transformers/gtr-t5-xl', 'gtr-t5-xxl': 'sentence-transformers/gtr-t5-xxl', 'sentence-t5-base': 'sentence-transformers/sentence-t5-base', 'sentence-t5-xl': 'sentence-transformers/sentence-t5-xl', 'sentence-t5-xxl': 'sentence-transformers/sentence-t5-xxl', 'sentence-t5-large': 'sentence-transformers/sentence-t5-large', 'distilbert-multilingual-nli-stsb-quora-ranking': 'sentence-transformers/distilbert-multilingual-nli-stsb-quora-ranking', 'msmarco-bert-co-condensor': 'sentence-transformers/msmarco-bert-co-condensor', 'msmarco-roberta-base-v2': 'sentence-transformers/msmarco-roberta-base-v2'}, 'colbert_ranker': {'colbertv2.0': 'colbert-ir/colbertv2.0', 'FranchColBERT': 'bclavie/FraColBERTv2', 'JapanColBERT': 'bclavie/JaColBERTv2', 'SpanishColBERT': 'AdrienB134/ColBERTv2.0-spanish-mmarcoES', 'jina-colbert-v1-en': 'jinaai/jina-colbert-v1-en', 'ArabicColBERT-250k': 'akhooli/arabic-colbertv2-250k-norm', 'ArabicColBERT-711k': 'akhooli/arabic-colbertv2-711k-norm', 'BengaliColBERT': 'turjo4nis/colbertv2.0-bn', 'mxbai-colbert-large-v1': 'mixedbread-ai/mxbai-colbert-large-v1'}, 'monobert': {'monobert-large': 'castorini/monobert-large-msmarco'}, 'llm2vec': {'Meta-Llama-31-8B': 'McGill-NLP/LLM2Vec-Meta-Llama-31-8B-Instruct-mntp', 'Meta-Llama-3-8B': 'McGill-NLP/LLM2Vec-Meta-Llama-3-8B-Instruct-mntp', 'Mistral-7B': 'McGill-NLP/LLM2Vec-Mistral-7B-Instruct-v2-mntp', 'Llama-2-7B': 'McGill-NLP/LLM2Vec-Llama-2-7b-chat-hf-mntp', 'Sheared-LLaMA': 'McGill-NLP/LLM2Vec-Sheared-LLaMA-mntp'}, 'twolar': {'twolar-xl': 'Dundalia/TWOLAR-xl', 'twolar-large': 'Dundalia/TWOLAR-large'}, 'duot5': {'duot5-base-msmarco': 'castorini/duot5-base-msmarco', 'duot5-3b-msmarco': 'castorini/duot5-3b-msmarco', 'duot5-large-msmarco': 'castorini/duot5-large-msmarco-10k'}, 'rankllama': {'rankllama-v1-7b-lora-passage': 'castorini/rankllama-v1-7b-lora-passage'}, 'dear_reranker': {'dear-3b-reranker-ce-v1': 'abdoelsayed/dear-3b-reranker-ce-v1', 'dear-3b-reranker-ranknet-v1': 'abdoelsayed/dear-3b-reranker-ranknet-v1', 'dear-3b-reranker-ce-lora-v1': 'abdoelsayed/dear-3b-reranker-ce-lora-v1', 'dear-8b-reranker-ce-v1': 'abdoelsayed/dear-8b-reranker-ce-v1'}, 'echorank': {'flan-t5-large': 'google/flan-t5-large', 'flan-t5-xl': 'google/flan-t5-xl'}, 'incontext_reranker': {'llamav3.1-8b': 'meta-llama/Meta-Llama-3.1-8B-Instruct', 'llamav3.1-70b': 'meta-llama/Llama-3.1-70B-Instruct', 'Mistral-7B-Instruct-v0.2': 'mistralai/Mistral-7B-Instruct-v0.2'}, 'tart': {'tart-full-flan-t5-xl': 'facebook/tart-full-flan-t5-xl', 'tart-dual-flan-t5-xl': 'facebook/tart-dual-flan-t5-xl'}, 'prp': {'llamav3.1-8b': 'meta-llama/Meta-Llama-3.1-8B-Instruct', 'llamav3.1-70b': 'meta-llama/Llama-3.1-70B-Instruct', 'Llama-3.2-1B': 'meta-llama/Llama-3.2-1B-Instruct', 'Llama-3.2-3B': 'meta-llama/Llama-3.2-3B-Instruct', 'Mistral-7B-Instruct-v0.3': 'mistralai/Mistral-7B-Instruct-v0.3'}, 'prp-api': {'gpt-3.5': 'gpt-3.5-turbo', 'gpt-4': 'gpt-4o', 'gpt-4-mini': 'gpt-4o-mini', 'llamav3.1-8b': 'llamav3.1-8b', 'llamav3.1-70b': 'llamav3.1-70b'}, 'rankgemma': {'gemma-2-2b': 'google/gemma-2-2b-it', 'gemma-2-9b': 'google/gemma-2-9b-it', 'gemma-2-27b': 'google/gemma-2-27b-it'}, 'rankmistral': {'mistral-7b': 'mistralai/Mistral-7B-Instruct-v0.3', 'mistral-7b-v0.2': 'mistralai/Mistral-7B-Instruct-v0.2', 'mixtral-8x7b': 'mistralai/Mixtral-8x7B-Instruct-v0.1'}} module-attribute

METHOD_MAP = {'upr': UPR, 'rankgpt-api': RankGPT, 'rankgpt': RankGPT, 'flashrank': FlashRanker, 'monot5': MonoT5, 'rankt5': RankT5, 'listt5': ListT5, 'inranker': InRanker, 'apiranker': APIRanker, 'transformer_ranker': TransformerRanker, 'llm_layerwise_ranker': LLMLayerWiseRanker, 'blender_reranker': BlenderReranker, 'splade_reranker': SpladeReranker, 'sentence_transformer_reranker': SentenceTransformerReranker, 'colbert_ranker': ColBERTReranker, 'monobert': MonoBERT, 'llm2vec': LLM2VecReranker, 'twolar': TWOLAR, 'echorank': EchoRankReranker, 'duot5': DuoT5, 'rankllama': RankLLaMAReranker, 'dear_reranker': DeARReranker, 'incontext_reranker': InContextReranker, 'prp': PRPReranker, 'prp-api': PRPReranker, 'tart': TARTReranker, 'rankgemma': RankGemmaReranker, 'rankmistral': RankMistralReranker} module-attribute

Document

Represents a document consisting of a question, answers, and contexts.

Attributes:

Name Type Description
question Question

The question associated with the document.
answers Answer

The answers to the question.
contexts list[Context]

A list of related contexts.
reorder_contexts list[Context] or None

A reordered list of contexts based on relevance.
Source code in rankify/dataset/dataset.py 
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
class Document:
    """
    Represents a document consisting of a question, answers, and contexts.

    Attributes:
        question (Question): The question associated with the document.
        answers (Answer): The answers to the question.
        contexts (list[Context]): A list of related contexts.
        reorder_contexts (list[Context] or None): A reordered list of contexts based on relevance.
    """
    def __init__(self, question: Question, answers: Answer, contexts: list = None , id: int = None) -> None:
        """
        Initializes a Document instance.

        Args:
            question (Question): The question associated with the document.
            answers (Answer): The answers to the question.
            contexts (list[Context], optional): A list of contexts related to the question.

        Example:
            ```python
            q = Question("What is the capital of France?")
            a = Answer(["Paris"])
            c1 = Context(score=0.9, has_answer=True, id=1, title="Paris", text="The capital of France is Paris.")
            c2 = Context(score=0.5, has_answer=False, id=2, title="Berlin", text="Berlin is the capital of Germany.")
            d = Document(question=q, answers=a, contexts=[c1, c2])
            print(d)
            ```
        """
        self.question: Question = question
        self.answers: Answer = answers
        self.contexts: List[Context] = contexts
        self.reorder_contexts: List[Context] = None
        self.id = str(id) 

    @classmethod
    def from_dict(cls, data: dict,n_docs:int=100) -> 'Document':
        """
        Creates a Document instance from a dictionary.

        Args:
            data (dict): A dictionary containing the question, answers, and contexts.
            n_docs (int, optional): The number of contexts to include. Defaults to 100.

        Returns:
            Document: A new Document instance.

        Example:
            ```python
            data = {
                "question": "What is the capital of France?",
                "answers": ["Paris"],
                "ctxs": [
                    {"score": 0.9, "has_answer": True, "id": 1, "title": "Paris", "text": "The capital of France is Paris."},
                    {"score": 0.5, "has_answer": False, "id": 2, "title": "Berlin", "text": "Berlin is the capital of Germany."}
                ]
            }
            d = Document.from_dict(data)
            print(d.question)
            ```
        """
        question = Question(data["question"])
        if "answers" in data:
            answers = Answer(data["answers"])
        else:
            answers =Answer('')

        if "query_id" in data:
            id = data["query_id"]
        else:
            id = None
        contexts = [Context(**ctx) for ctx in data["ctxs"][:n_docs]]
        return cls(question, answers, contexts, id=id)

    def to_dict(self) -> Dict[str, Optional[object]]:
        """
        Converts the document into a dictionary representation.

        Returns:
            dict: A dictionary containing the question, answers, and contexts.
        """
        return {
            "question": self.question.question,
            "answers": self.answers.answers,
            "contexts": [ctx.to_dict() for ctx in self.contexts]
        }
    def to_dict_reoreder(self) -> Dict[str,Optional[object]]:
        return {
            "question" : self.question.question,
            "answers" : self.answers.answers,
            "contexts" : [ctx.to_dict() for ctx in self.reorder_contexts]
        }
    def __str__(self) -> str:
        """
        Returns a string representation of the Document instance.

        Returns:
            str: The formatted document information.

        Example:
            ```python
            d = Document(Question("What is the capital of France?"), Answer(["Paris"]))
            print(d)
            ```
        """
        contexts_str = "\n\n".join([str(ctx) for ctx in self.contexts])
        reorder_contexts_str= ''
        if self.reorder_contexts is not None:
            reorder_contexts_str = "\n\n".join([str(ctx) for ctx in self.reorder_contexts])
        return f"{self.question}\n\n{self.answers}\n\nContext: \n\n{contexts_str}\nReorder contexts: \n\n{reorder_contexts_str}"

__init__(question, answers, contexts=None, id=None)

Initializes a Document instance.

Parameters:

Name Type Description Default
question Question

The question associated with the document.

 required
answers Answer

The answers to the question.

 required
contexts list[Context]

A list of contexts related to the question.

 None
Example 
q = Question("What is the capital of France?")
a = Answer(["Paris"])
c1 = Context(score=0.9, has_answer=True, id=1, title="Paris", text="The capital of France is Paris.")
c2 = Context(score=0.5, has_answer=False, id=2, title="Berlin", text="Berlin is the capital of Germany.")
d = Document(question=q, answers=a, contexts=[c1, c2])
print(d)
Source code in rankify/dataset/dataset.py 
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
def __init__(self, question: Question, answers: Answer, contexts: list = None , id: int = None) -> None:
    """
    Initializes a Document instance.

    Args:
        question (Question): The question associated with the document.
        answers (Answer): The answers to the question.
        contexts (list[Context], optional): A list of contexts related to the question.

    Example:
        ```python
        q = Question("What is the capital of France?")
        a = Answer(["Paris"])
        c1 = Context(score=0.9, has_answer=True, id=1, title="Paris", text="The capital of France is Paris.")
        c2 = Context(score=0.5, has_answer=False, id=2, title="Berlin", text="Berlin is the capital of Germany.")
        d = Document(question=q, answers=a, contexts=[c1, c2])
        print(d)
        ```
    """
    self.question: Question = question
    self.answers: Answer = answers
    self.contexts: List[Context] = contexts
    self.reorder_contexts: List[Context] = None
    self.id = str(id)

from_dict(data, n_docs=100) classmethod

Creates a Document instance from a dictionary.

Parameters:

Name Type Description Default
data dict

A dictionary containing the question, answers, and contexts.

 required
n_docs int

The number of contexts to include. Defaults to 100.

 100

Returns:

Name Type Description
Document Document

A new Document instance.
Example 
data = {
    "question": "What is the capital of France?",
    "answers": ["Paris"],
    "ctxs": [
        {"score": 0.9, "has_answer": True, "id": 1, "title": "Paris", "text": "The capital of France is Paris."},
        {"score": 0.5, "has_answer": False, "id": 2, "title": "Berlin", "text": "Berlin is the capital of Germany."}
    ]
}
d = Document.from_dict(data)
print(d.question)
Source code in rankify/dataset/dataset.py 
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
@classmethod
def from_dict(cls, data: dict,n_docs:int=100) -> 'Document':
    """
    Creates a Document instance from a dictionary.

    Args:
        data (dict): A dictionary containing the question, answers, and contexts.
        n_docs (int, optional): The number of contexts to include. Defaults to 100.

    Returns:
        Document: A new Document instance.

    Example:
        ```python
        data = {
            "question": "What is the capital of France?",
            "answers": ["Paris"],
            "ctxs": [
                {"score": 0.9, "has_answer": True, "id": 1, "title": "Paris", "text": "The capital of France is Paris."},
                {"score": 0.5, "has_answer": False, "id": 2, "title": "Berlin", "text": "Berlin is the capital of Germany."}
            ]
        }
        d = Document.from_dict(data)
        print(d.question)
        ```
    """
    question = Question(data["question"])
    if "answers" in data:
        answers = Answer(data["answers"])
    else:
        answers =Answer('')

    if "query_id" in data:
        id = data["query_id"]
    else:
        id = None
    contexts = [Context(**ctx) for ctx in data["ctxs"][:n_docs]]
    return cls(question, answers, contexts, id=id)

to_dict()

Converts the document into a dictionary representation.

Returns:

Name Type Description
dict Dict[str, Optional[object]]

A dictionary containing the question, answers, and contexts.
Source code in rankify/dataset/dataset.py 
269
270
271
272
273
274
275
276
277
278
279
280
def to_dict(self) -> Dict[str, Optional[object]]:
    """
    Converts the document into a dictionary representation.

    Returns:
        dict: A dictionary containing the question, answers, and contexts.
    """
    return {
        "question": self.question.question,
        "answers": self.answers.answers,
        "contexts": [ctx.to_dict() for ctx in self.contexts]
    }

__str__()

Returns a string representation of the Document instance.

Returns:

Name Type Description
str str

The formatted document information.
Example 
d = Document(Question("What is the capital of France?"), Answer(["Paris"]))
print(d)
Source code in rankify/dataset/dataset.py 
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
def __str__(self) -> str:
    """
    Returns a string representation of the Document instance.

    Returns:
        str: The formatted document information.

    Example:
        ```python
        d = Document(Question("What is the capital of France?"), Answer(["Paris"]))
        print(d)
        ```
    """
    contexts_str = "\n\n".join([str(ctx) for ctx in self.contexts])
    reorder_contexts_str= ''
    if self.reorder_contexts is not None:
        reorder_contexts_str = "\n\n".join([str(ctx) for ctx in self.reorder_contexts])
    return f"{self.question}\n\n{self.answers}\n\nContext: \n\n{contexts_str}\nReorder contexts: \n\n{reorder_contexts_str}"

Reranking

A class for reranking documents using different ranking models.

Attributes:

Name Type Description
method str

The name of the reranking method to be used.
model str

The name of the model used for reranking.
api_key str

An optional API key for accessing remote models or services.
ranker BaseRanking

An instance of a ranking model used for reranking.
Source code in rankify/models/reranking.py 
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
class Reranking:
    """
    A class for reranking documents using different ranking models.

    Attributes:
        method (str): The name of the reranking method to be used.
        model (str): The name of the model used for reranking.
        api_key (str, optional): An optional API key for accessing remote models or services.
        ranker (BaseRanking): An instance of a ranking model used for reranking.
    """
    def __init__(self, method: str= None, model_name: str= None, api_key: str = None , **kwargs):
        """
        Initializes a Reranking instance.

        Args:
            method (str, optional): The name of the reranking method to be used. Defaults to None.
            model_name (str, optional): The name of the model to be used for reranking. Defaults to None.
            api_key (str, optional): An optional API key for accessing remote models or services. Defaults to None.
            **kwargs: Additional arguments passed to the reranking model.

        Example:
            ```python
            model = Reranking(method="listt5", model_name="listt5-base")
            ```
        """
        self.method = method
        self.model = model_name
        self.api_key= api_key
        self.kwargs = kwargs
        self.ranker = self.initialize()

    def initialize(self):
        """
        Initializes the ranking model based on the specified method and model name.

        Returns:
            BaseRanking: An instance of a ranking model.

        Raises:
            ValueError: If the specified ranking method is not supported.

        Example:
            ```python
            model = Reranking(method="listt5", model_name="listt5-base")
            ranker = model.initialize()
            ```
        """

        if self.model not in HF_PRE_DEFIND_MODELS[self.method]:
            model_name= self.model
        else:
            model_name= HF_PRE_DEFIND_MODELS[self.method][self.model]

        if self.method in METHOD_MAP:
            return METHOD_MAP[self.method](method=self.method,model_name=model_name, api_key=self.api_key, **self.kwargs)
        else:
            raise ValueError(f"Ranking method {self.method} is not supported or installed. please install the full version of Rankify")

    def rank(self, documents: list[Document]):
        """
        Reranks a list of documents using the specified ranking model.

        Args:
            documents (list[Document]): A list of Document instances that need to be reranked.

        Returns:
            list[Document]: The reranked list of Document instances.

        Example:
            ```python
            question = Question("When did Thomas Edison invent the light bulb?")
            answers = Answer(["1879"])
            contexts = [
                Context(text="Lightning strike at Seoul National University", id=1),
                Context(text="Thomas Edison invented the light bulb in 1879", id=5)
            ]
            document = Document(question=question, answers=answers, contexts=contexts)
            model = Reranking(method="listt5", model_name="listt5-base")
            model.rank([document])
            ```
        """
        return self.ranker.rank(documents=documents)

__init__(method=None, model_name=None, api_key=None, **kwargs)

Initializes a Reranking instance.

Parameters:

Name Type Description Default
method str

The name of the reranking method to be used. Defaults to None.

 None
model_name str

The name of the model to be used for reranking. Defaults to None.

 None
api_key str

An optional API key for accessing remote models or services. Defaults to None.

 None
**kwargs

Additional arguments passed to the reranking model.

 {}
Example 
model = Reranking(method="listt5", model_name="listt5-base")
Source code in rankify/models/reranking.py 
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
def __init__(self, method: str= None, model_name: str= None, api_key: str = None , **kwargs):
    """
    Initializes a Reranking instance.

    Args:
        method (str, optional): The name of the reranking method to be used. Defaults to None.
        model_name (str, optional): The name of the model to be used for reranking. Defaults to None.
        api_key (str, optional): An optional API key for accessing remote models or services. Defaults to None.
        **kwargs: Additional arguments passed to the reranking model.

    Example:
        ```python
        model = Reranking(method="listt5", model_name="listt5-base")
        ```
    """
    self.method = method
    self.model = model_name
    self.api_key= api_key
    self.kwargs = kwargs
    self.ranker = self.initialize()

initialize()

Initializes the ranking model based on the specified method and model name.

Returns:

Name Type Description
BaseRanking

An instance of a ranking model.

Raises:

Type Description
ValueError

If the specified ranking method is not supported.
Example 
model = Reranking(method="listt5", model_name="listt5-base")
ranker = model.initialize()
Source code in rankify/models/reranking.py 
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
def initialize(self):
    """
    Initializes the ranking model based on the specified method and model name.

    Returns:
        BaseRanking: An instance of a ranking model.

    Raises:
        ValueError: If the specified ranking method is not supported.

    Example:
        ```python
        model = Reranking(method="listt5", model_name="listt5-base")
        ranker = model.initialize()
        ```
    """

    if self.model not in HF_PRE_DEFIND_MODELS[self.method]:
        model_name= self.model
    else:
        model_name= HF_PRE_DEFIND_MODELS[self.method][self.model]

    if self.method in METHOD_MAP:
        return METHOD_MAP[self.method](method=self.method,model_name=model_name, api_key=self.api_key, **self.kwargs)
    else:
        raise ValueError(f"Ranking method {self.method} is not supported or installed. please install the full version of Rankify")

rank(documents)

Reranks a list of documents using the specified ranking model.

Parameters:

Name Type Description Default
documents list[Document]

A list of Document instances that need to be reranked.

 required

Returns:

Type Description

list[Document]: The reranked list of Document instances.
Example 
question = Question("When did Thomas Edison invent the light bulb?")
answers = Answer(["1879"])
contexts = [
    Context(text="Lightning strike at Seoul National University", id=1),
    Context(text="Thomas Edison invented the light bulb in 1879", id=5)
]
document = Document(question=question, answers=answers, contexts=contexts)
model = Reranking(method="listt5", model_name="listt5-base")
model.rank([document])
Source code in rankify/models/reranking.py 
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
def rank(self, documents: list[Document]):
    """
    Reranks a list of documents using the specified ranking model.

    Args:
        documents (list[Document]): A list of Document instances that need to be reranked.

    Returns:
        list[Document]: The reranked list of Document instances.

    Example:
        ```python
        question = Question("When did Thomas Edison invent the light bulb?")
        answers = Answer(["1879"])
        contexts = [
            Context(text="Lightning strike at Seoul National University", id=1),
            Context(text="Thomas Edison invented the light bulb in 1879", id=5)
        ]
        document = Document(question=question, answers=answers, contexts=contexts)
        model = Reranking(method="listt5", model_name="listt5-base")
        model.rank([document])
        ```
    """
    return self.ranker.rank(documents=documents)