from __future__ import annotations import copy import importlib import inspect import json import logging import math import os import queue import shutil import sys import tempfile import traceback import warnings from collections import OrderedDict from collections.abc import Iterable, Iterator from contextlib import contextmanager from multiprocessing import Queue from pathlib import Path from typing import Any, Callable, Literal, overload import numpy as np import numpy.typing as npt import torch import torch.multiprocessing as mp import transformers from huggingface_hub import HfApi from packaging import version from torch import Tensor, device, nn from tqdm.autonotebook import trange from transformers import PreTrainedModel, is_torch_npu_available from transformers.dynamic_module_utils import get_class_from_dynamic_module, get_relative_import_files from typing_extensions import deprecated from sentence_transformers.model_card import SentenceTransformerModelCardData, generate_model_card from sentence_transformers.models import Router from sentence_transformers.models.Module import Module from sentence_transformers.similarity_functions import SimilarityFunction from . import __MODEL_HUB_ORGANIZATION__, __version__ from .evaluation import SentenceEvaluator from .fit_mixin import FitMixin from .models import Pooling, Transformer from .peft_mixin import PeftAdapterMixin from .quantization import quantize_embeddings from .util import ( batch_to_device, get_device_name, import_from_string, is_sentence_transformer_model, load_dir_path, load_file_path, save_to_hub_args_decorator, truncate_embeddings, ) logger = logging.getLogger(__name__) class SentenceTransformer(nn.Sequential, FitMixin, PeftAdapterMixin): """ Loads or creates a SentenceTransformer model that can be used to map sentences / text to embeddings. Args: model_name_or_path (str, optional): If it is a filepath on disc, it loads the model from that path. If it is not a path, it first tries to download a pre-trained SentenceTransformer model. If that fails, tries to construct a model from the Hugging Face Hub with that name. modules (Iterable[nn.Module], optional): A list of torch Modules that should be called sequentially, can be used to create custom SentenceTransformer models from scratch. device (str, optional): Device (like "cuda", "cpu", "mps", "npu") that should be used for computation. If None, checks if a GPU can be used. prompts (Dict[str, str], optional): A dictionary with prompts for the model. The key is the prompt name, the value is the prompt text. The prompt text will be prepended before any text to encode. For example: `{"query": "query: ", "passage": "passage: "}` or `{"clustering": "Identify the main category based on the titles in "}`. default_prompt_name (str, optional): The name of the prompt that should be used by default. If not set, no prompt will be applied. similarity_fn_name (str or SimilarityFunction, optional): The name of the similarity function to use. Valid options are "cosine", "dot", "euclidean", and "manhattan". If not set, it is automatically set to "cosine" if `similarity` or `similarity_pairwise` are called while `model.similarity_fn_name` is still `None`. cache_folder (str, optional): Path to store models. Can also be set by the SENTENCE_TRANSFORMERS_HOME environment variable. trust_remote_code (bool, optional): Whether or not to allow for custom models defined on the Hub in their own modeling files. This option should only be set to True for repositories you trust and in which you have read the code, as it will execute code present on the Hub on your local machine. revision (str, optional): The specific model version to use. It can be a branch name, a tag name, or a commit id, for a stored model on Hugging Face. local_files_only (bool, optional): Whether or not to only look at local files (i.e., do not try to download the model). token (bool or str, optional): Hugging Face authentication token to download private models. use_auth_token (bool or str, optional): Deprecated argument. Please use `token` instead. truncate_dim (int, optional): The dimension to truncate sentence embeddings to. Defaults to None. model_kwargs (Dict[str, Any], optional): Additional model configuration parameters to be passed to the Hugging Face Transformers model. Particularly useful options are: - ``torch_dtype``: Override the default `torch.dtype` and load the model under a specific `dtype`. The different options are: 1. ``torch.float16``, ``torch.bfloat16`` or ``torch.float``: load in a specified ``dtype``, ignoring the model's ``config.torch_dtype`` if one exists. If not specified - the model will get loaded in ``torch.float`` (fp32). 2. ``"auto"`` - A ``torch_dtype`` entry in the ``config.json`` file of the model will be attempted to be used. If this entry isn't found then next check the ``dtype`` of the first weight in the checkpoint that's of a floating point type and use that as ``dtype``. This will load the model using the ``dtype`` it was saved in at the end of the training. It can't be used as an indicator of how the model was trained. Since it could be trained in one of half precision dtypes, but saved in fp32. - ``attn_implementation``: The attention implementation to use in the model (if relevant). Can be any of `"eager"` (manual implementation of the attention), `"sdpa"` (using `F.scaled_dot_product_attention `_), or `"flash_attention_2"` (using `Dao-AILab/flash-attention `_). By default, if available, SDPA will be used for torch>=2.1.1. The default is otherwise the manual `"eager"` implementation. - ``provider``: If backend is "onnx", this is the provider to use for inference, for example "CPUExecutionProvider", "CUDAExecutionProvider", etc. See https://onnxruntime.ai/docs/execution-providers/ for all ONNX execution providers. - ``file_name``: If backend is "onnx" or "openvino", this is the file name to load, useful for loading optimized or quantized ONNX or OpenVINO models. - ``export``: If backend is "onnx" or "openvino", then this is a boolean flag specifying whether this model should be exported to the backend. If not specified, the model will be exported only if the model repository or directory does not already contain an exported model. See the `PreTrainedModel.from_pretrained `_ documentation for more details. tokenizer_kwargs (Dict[str, Any], optional): Additional tokenizer configuration parameters to be passed to the Hugging Face Transformers tokenizer. See the `AutoTokenizer.from_pretrained `_ documentation for more details. config_kwargs (Dict[str, Any], optional): Additional model configuration parameters to be passed to the Hugging Face Transformers config. See the `AutoConfig.from_pretrained `_ documentation for more details. model_card_data (:class:`~sentence_transformers.model_card.SentenceTransformerModelCardData`, optional): A model card data object that contains information about the model. This is used to generate a model card when saving the model. If not set, a default model card data object is created. backend (str): The backend to use for inference. Can be one of "torch" (default), "onnx", or "openvino". See https://sbert.net/docs/sentence_transformer/usage/efficiency.html for benchmarking information on the different backends. Example: :: from sentence_transformers import SentenceTransformer # Load a pre-trained SentenceTransformer model model = SentenceTransformer('all-mpnet-base-v2') # Encode some texts sentences = [ "The weather is lovely today.", "It's so sunny outside!", "He drove to the stadium.", ] embeddings = model.encode(sentences) print(embeddings.shape) # (3, 768) # Get the similarity scores between all sentences similarities = model.similarity(embeddings, embeddings) print(similarities) # tensor([[1.0000, 0.6817, 0.0492], # [0.6817, 1.0000, 0.0421], # [0.0492, 0.0421, 1.0000]]) """ model_card_data_class = SentenceTransformerModelCardData def __init__( self, model_name_or_path: str | None = None, modules: Iterable[nn.Module] | None = None, device: str | None = None, prompts: dict[str, str] | None = None, default_prompt_name: str | None = None, similarity_fn_name: str | SimilarityFunction | None = None, cache_folder: str | None = None, trust_remote_code: bool = False, revision: str | None = None, local_files_only: bool = False, token: bool | str | None = None, use_auth_token: bool | str | None = None, truncate_dim: int | None = None, model_kwargs: dict[str, Any] | None = None, tokenizer_kwargs: dict[str, Any] | None = None, config_kwargs: dict[str, Any] | None = None, model_card_data: SentenceTransformerModelCardData | None = None, backend: Literal["torch", "onnx", "openvino"] = "torch", ) -> None: # Note: self._load_sbert_model can also update `self.prompts` and `self.default_prompt_name` self.prompts = {"query": "", "document": ""} if prompts: self.prompts.update(prompts) self.default_prompt_name = default_prompt_name self.similarity_fn_name = similarity_fn_name self.trust_remote_code = trust_remote_code self.truncate_dim = truncate_dim self.model_card_data = model_card_data or self.model_card_data_class(local_files_only=local_files_only) self.module_kwargs = None self._model_card_vars = {} self._model_card_text = None self._model_config = {"model_type": self.__class__.__name__} self._prompt_length_mapping = {} self.backend = backend if use_auth_token is not None: warnings.warn( "The `use_auth_token` argument is deprecated and will be removed in v4 of SentenceTransformers.", FutureWarning, ) if token is not None: raise ValueError( "`token` and `use_auth_token` are both specified. Please set only the argument `token`." ) token = use_auth_token if cache_folder is None: cache_folder = os.getenv("SENTENCE_TRANSFORMERS_HOME") if device is None: device = get_device_name() logger.info(f"Use pytorch device_name: {device}") if device == "hpu" and importlib.util.find_spec("optimum") is not None: from optimum.habana.transformers.modeling_utils import adapt_transformers_to_gaudi adapt_transformers_to_gaudi() if model_name_or_path is not None and model_name_or_path != "": logger.info(f"Load pretrained {self.__class__.__name__}: {model_name_or_path}") # Old models that don't belong to any organization basic_transformer_models = [ "albert-base-v1", "albert-base-v2", "albert-large-v1", "albert-large-v2", "albert-xlarge-v1", "albert-xlarge-v2", "albert-xxlarge-v1", "albert-xxlarge-v2", "bert-base-cased-finetuned-mrpc", "bert-base-cased", "bert-base-chinese", "bert-base-german-cased", "bert-base-german-dbmdz-cased", "bert-base-german-dbmdz-uncased", "bert-base-multilingual-cased", "bert-base-multilingual-uncased", "bert-base-uncased", "bert-large-cased-whole-word-masking-finetuned-squad", "bert-large-cased-whole-word-masking", "bert-large-cased", "bert-large-uncased-whole-word-masking-finetuned-squad", "bert-large-uncased-whole-word-masking", "bert-large-uncased", "camembert-base", "ctrl", "distilbert-base-cased-distilled-squad", "distilbert-base-cased", "distilbert-base-german-cased", "distilbert-base-multilingual-cased", "distilbert-base-uncased-distilled-squad", "distilbert-base-uncased-finetuned-sst-2-english", "distilbert-base-uncased", "distilgpt2", "distilroberta-base", "gpt2-large", "gpt2-medium", "gpt2-xl", "gpt2", "openai-gpt", "roberta-base-openai-detector", "roberta-base", "roberta-large-mnli", "roberta-large-openai-detector", "roberta-large", "t5-11b", "t5-3b", "t5-base", "t5-large", "t5-small", "transfo-xl-wt103", "xlm-clm-ende-1024", "xlm-clm-enfr-1024", "xlm-mlm-100-1280", "xlm-mlm-17-1280", "xlm-mlm-en-2048", "xlm-mlm-ende-1024", "xlm-mlm-enfr-1024", "xlm-mlm-enro-1024", "xlm-mlm-tlm-xnli15-1024", "xlm-mlm-xnli15-1024", "xlm-roberta-base", "xlm-roberta-large-finetuned-conll02-dutch", "xlm-roberta-large-finetuned-conll02-spanish", "xlm-roberta-large-finetuned-conll03-english", "xlm-roberta-large-finetuned-conll03-german", "xlm-roberta-large", "xlnet-base-cased", "xlnet-large-cased", ] if not os.path.exists(model_name_or_path): # Not a path, load from hub if "\\" in model_name_or_path or model_name_or_path.count("/") > 1: raise FileNotFoundError(f"Path {model_name_or_path} not found") if "/" not in model_name_or_path and model_name_or_path.lower() not in basic_transformer_models: # A model from sentence-transformers model_name_or_path = __MODEL_HUB_ORGANIZATION__ + "/" + model_name_or_path has_modules = is_sentence_transformer_model( model_name_or_path, token, cache_folder=cache_folder, revision=revision, local_files_only=local_files_only, ) if ( has_modules and self._get_model_type( model_name_or_path, token, cache_folder=cache_folder, revision=revision, local_files_only=local_files_only, ) == self._model_config["model_type"] ): modules, self.module_kwargs = self._load_sbert_model( model_name_or_path, token=token, cache_folder=cache_folder, revision=revision, trust_remote_code=trust_remote_code, local_files_only=local_files_only, model_kwargs=model_kwargs, tokenizer_kwargs=tokenizer_kwargs, config_kwargs=config_kwargs, ) else: modules = self._load_auto_model( model_name_or_path, token=token, cache_folder=cache_folder, revision=revision, trust_remote_code=trust_remote_code, local_files_only=local_files_only, model_kwargs=model_kwargs, tokenizer_kwargs=tokenizer_kwargs, config_kwargs=config_kwargs, has_modules=has_modules, ) if modules is not None and not isinstance(modules, OrderedDict): modules = OrderedDict([(str(idx), module) for idx, module in enumerate(modules)]) super().__init__(modules) # Ensure all tensors in the model are of the same dtype as the first tensor # This is necessary if the first module has been given a lower precision via # model_kwargs["torch_dtype"]. The rest of the model should be loaded in the same dtype # See #2887 for more details try: dtype = next(self.parameters()).dtype self.to(dtype) except StopIteration: pass self.to(device) self.is_hpu_graph_enabled = False if self.default_prompt_name is not None and self.default_prompt_name not in self.prompts: raise ValueError( f"Default prompt name '{self.default_prompt_name}' not found in the configured prompts " f"dictionary with keys {list(self.prompts.keys())!r}." ) if self.prompts and (non_empty_keys := [k for k, v in self.prompts.items() if v != ""]): if len(non_empty_keys) == 1: logger.info(f"1 prompt is loaded, with the key: {non_empty_keys[0]}") else: logger.info(f"{len(non_empty_keys)} prompts are loaded, with the keys: {non_empty_keys}") if self.default_prompt_name: logger.warning( f"Default prompt name is set to '{self.default_prompt_name}'. " "This prompt will be applied to all `encode()` calls, except if `encode()` " "is called with `prompt` or `prompt_name` parameters." ) # Ideally, INSTRUCTOR models should set `include_prompt=False` in their pooling configuration, but # that would be a breaking change for users currently using the InstructorEmbedding project. # So, instead we hardcode setting it for the main INSTRUCTOR models, and otherwise give a warning if we # suspect the user is using an INSTRUCTOR model. if model_name_or_path in ("hkunlp/instructor-base", "hkunlp/instructor-large", "hkunlp/instructor-xl"): self.set_pooling_include_prompt(include_prompt=False) elif ( model_name_or_path and "/" in model_name_or_path and "instructor" in model_name_or_path.split("/")[1].lower() ): if any([module.include_prompt for module in self if isinstance(module, Pooling)]): logger.warning( "Instructor models require `include_prompt=False` in the pooling configuration. " "Either update the model configuration or call `model.set_pooling_include_prompt(False)` after loading the model." ) # Pass the model to the model card data for later use in generating a model card upon saving this model self.model_card_data.register_model(self) def get_backend(self) -> Literal["torch", "onnx", "openvino"]: """Return the backend used for inference, which can be one of "torch", "onnx", or "openvino". Returns: str: The backend used for inference. """ return self.backend def encode_query( self, sentences: str | list[str] | np.ndarray, prompt_name: str | None = None, prompt: str | None = None, batch_size: int = 32, show_progress_bar: bool | None = None, output_value: Literal["sentence_embedding", "token_embeddings"] | None = "sentence_embedding", precision: Literal["float32", "int8", "uint8", "binary", "ubinary"] = "float32", convert_to_numpy: bool = True, convert_to_tensor: bool = False, device: str | list[str | torch.device] | None = None, normalize_embeddings: bool = False, truncate_dim: int | None = None, pool: dict[Literal["input", "output", "processes"], Any] | None = None, chunk_size: int | None = None, **kwargs, ) -> list[Tensor] | np.ndarray | Tensor | dict[str, Tensor] | list[dict[str, Tensor]]: """ Computes sentence embeddings specifically optimized for query representation. This method is a specialized version of :meth:`encode` that differs in exactly two ways: 1. If no ``prompt_name`` or ``prompt`` is provided, it uses a predefined "query" prompt, if available in the model's ``prompts`` dictionary. 2. It sets the ``task`` to "query". If the model has a :class:`~sentence_transformers.models.Router` module, it will use the "query" task type to route the input through the appropriate submodules. .. tip:: If you are unsure whether you should use :meth:`encode`, :meth:`encode_query`, or :meth:`encode_document`, your best bet is to use :meth:`encode_query` and :meth:`encode_document` for Information Retrieval tasks with clear query and document/passage distinction, and use :meth:`encode` for all other tasks. Note that :meth:`encode` is the most general method and can be used for any task, including Information Retrieval, and that if the model was not trained with predefined prompts and/or task types, then all three methods will return identical embeddings. Args: sentences (Union[str, List[str]]): The sentences to embed. prompt_name (Optional[str], optional): The name of the prompt to use for encoding. Must be a key in the `prompts` dictionary, which is either set in the constructor or loaded from the model configuration. For example if ``prompt_name`` is "query" and the ``prompts`` is {"query": "query: ", ...}, then the sentence "What is the capital of France?" will be encoded as "query: What is the capital of France?" because the sentence is appended to the prompt. If ``prompt`` is also set, this argument is ignored. Defaults to None. prompt (Optional[str], optional): The prompt to use for encoding. For example, if the prompt is "query: ", then the sentence "What is the capital of France?" will be encoded as "query: What is the capital of France?" because the sentence is appended to the prompt. If ``prompt`` is set, ``prompt_name`` is ignored. Defaults to None. batch_size (int, optional): The batch size used for the computation. Defaults to 32. show_progress_bar (bool, optional): Whether to output a progress bar when encode sentences. Defaults to None. output_value (Optional[Literal["sentence_embedding", "token_embeddings"]], optional): The type of embeddings to return: "sentence_embedding" to get sentence embeddings, "token_embeddings" to get wordpiece token embeddings, and `None`, to get all output values. Defaults to "sentence_embedding". precision (Literal["float32", "int8", "uint8", "binary", "ubinary"], optional): The precision to use for the embeddings. Can be "float32", "int8", "uint8", "binary", or "ubinary". All non-float32 precisions are quantized embeddings. Quantized embeddings are smaller in size and faster to compute, but may have a lower accuracy. They are useful for reducing the size of the embeddings of a corpus for semantic search, among other tasks. Defaults to "float32". convert_to_numpy (bool, optional): Whether the output should be a list of numpy vectors. If False, it is a list of PyTorch tensors. Defaults to True. convert_to_tensor (bool, optional): Whether the output should be one large tensor. Overwrites `convert_to_numpy`. Defaults to False. device (Union[str, List[str], None], optional): Device(s) to use for computation. Can be: - A single device string (e.g., "cuda:0", "cpu") for single-process encoding - A list of device strings (e.g., ["cuda:0", "cuda:1"], ["cpu", "cpu", "cpu", "cpu"]) to distribute encoding across multiple processes - None to auto-detect available device for single-process encoding If a list is provided, multi-process encoding will be used. Defaults to None. normalize_embeddings (bool, optional): Whether to normalize returned vectors to have length 1. In that case, the faster dot-product (util.dot_score) instead of cosine similarity can be used. Defaults to False. truncate_dim (int, optional): The dimension to truncate sentence embeddings to. Truncation is especially interesting for `Matryoshka models `_, i.e. models that are trained to still produce useful embeddings even if the embedding dimension is reduced. Truncated embeddings require less memory and are faster to perform retrieval with, but note that inference is just as fast, and the embedding performance is worse than the full embeddings. If None, the ``truncate_dim`` from the model initialization is used. Defaults to None. pool (Dict[Literal["input", "output", "processes"], Any], optional): A pool created by `start_multi_process_pool()` for multi-process encoding. If provided, the encoding will be distributed across multiple processes. This is recommended for large datasets and when multiple GPUs are available. Defaults to None. chunk_size (int, optional): Size of chunks for multi-process encoding. Only used with multiprocessing, i.e. when ``pool`` is not None or ``device`` is a list. If None, a sensible default is calculated. Defaults to None. Returns: Union[List[Tensor], ndarray, Tensor]: By default, a 2d numpy array with shape [num_inputs, output_dimension] is returned. If only one string input is provided, then the output is a 1d array with shape [output_dimension]. If ``convert_to_tensor``, a torch Tensor is returned instead. If ``self.truncate_dim <= output_dimension`` then output_dimension is ``self.truncate_dim``. Example: :: from sentence_transformers import SentenceTransformer # Load a pre-trained SentenceTransformer model model = SentenceTransformer("mixedbread-ai/mxbai-embed-large-v1") # Encode some queries queries = [ "What are the effects of climate change?", "History of artificial intelligence", "Technical specifications product XYZ", ] # Using query-specific encoding embeddings = model.encode_query(queries) print(embeddings.shape) # (3, 768) """ if prompt_name is None and "query" in self.prompts and prompt is None: prompt_name = "query" return self.encode( sentences=sentences, prompt_name=prompt_name, prompt=prompt, batch_size=batch_size, show_progress_bar=show_progress_bar, output_value=output_value, precision=precision, convert_to_numpy=convert_to_numpy, convert_to_tensor=convert_to_tensor, device=device, normalize_embeddings=normalize_embeddings, truncate_dim=truncate_dim, pool=pool, chunk_size=chunk_size, task="query", **kwargs, ) def encode_document( self, sentences: str | list[str] | np.ndarray, prompt_name: str | None = None, prompt: str | None = None, batch_size: int = 32, show_progress_bar: bool | None = None, output_value: Literal["sentence_embedding", "token_embeddings"] | None = "sentence_embedding", precision: Literal["float32", "int8", "uint8", "binary", "ubinary"] = "float32", convert_to_numpy: bool = True, convert_to_tensor: bool = False, device: str | list[str | torch.device] | None = None, normalize_embeddings: bool = False, truncate_dim: int | None = None, pool: dict[Literal["input", "output", "processes"], Any] | None = None, chunk_size: int | None = None, **kwargs, ) -> list[Tensor] | np.ndarray | Tensor | dict[str, Tensor] | list[dict[str, Tensor]]: """ Computes sentence embeddings specifically optimized for document/passage representation. This method is a specialized version of :meth:`encode` that differs in exactly two ways: 1. If no ``prompt_name`` or ``prompt`` is provided, it uses a predefined "document" prompt, if available in the model's ``prompts`` dictionary. 2. It sets the ``task`` to "document". If the model has a :class:`~sentence_transformers.models.Router` module, it will use the "document" task type to route the input through the appropriate submodules. .. tip:: If you are unsure whether you should use :meth:`encode`, :meth:`encode_query`, or :meth:`encode_document`, your best bet is to use :meth:`encode_query` and :meth:`encode_document` for Information Retrieval tasks with clear query and document/passage distinction, and use :meth:`encode` for all other tasks. Note that :meth:`encode` is the most general method and can be used for any task, including Information Retrieval, and that if the model was not trained with predefined prompts and/or task types, then all three methods will return identical embeddings. Args: sentences (Union[str, List[str]]): The sentences to embed. prompt_name (Optional[str], optional): The name of the prompt to use for encoding. Must be a key in the `prompts` dictionary, which is either set in the constructor or loaded from the model configuration. For example if ``prompt_name`` is "query" and the ``prompts`` is {"query": "query: ", ...}, then the sentence "What is the capital of France?" will be encoded as "query: What is the capital of France?" because the sentence is appended to the prompt. If ``prompt`` is also set, this argument is ignored. Defaults to None. prompt (Optional[str], optional): The prompt to use for encoding. For example, if the prompt is "query: ", then the sentence "What is the capital of France?" will be encoded as "query: What is the capital of France?" because the sentence is appended to the prompt. If ``prompt`` is set, ``prompt_name`` is ignored. Defaults to None. batch_size (int, optional): The batch size used for the computation. Defaults to 32. show_progress_bar (bool, optional): Whether to output a progress bar when encode sentences. Defaults to None. output_value (Optional[Literal["sentence_embedding", "token_embeddings"]], optional): The type of embeddings to return: "sentence_embedding" to get sentence embeddings, "token_embeddings" to get wordpiece token embeddings, and `None`, to get all output values. Defaults to "sentence_embedding". precision (Literal["float32", "int8", "uint8", "binary", "ubinary"], optional): The precision to use for the embeddings. Can be "float32", "int8", "uint8", "binary", or "ubinary". All non-float32 precisions are quantized embeddings. Quantized embeddings are smaller in size and faster to compute, but may have a lower accuracy. They are useful for reducing the size of the embeddings of a corpus for semantic search, among other tasks. Defaults to "float32". convert_to_numpy (bool, optional): Whether the output should be a list of numpy vectors. If False, it is a list of PyTorch tensors. Defaults to True. convert_to_tensor (bool, optional): Whether the output should be one large tensor. Overwrites `convert_to_numpy`. Defaults to False. device (Union[str, List[str], None], optional): Device(s) to use for computation. Can be: - A single device string (e.g., "cuda:0", "cpu") for single-process encoding - A list of device strings (e.g., ["cuda:0", "cuda:1"], ["cpu", "cpu", "cpu", "cpu"]) to distribute encoding across multiple processes - None to auto-detect available device for single-process encoding If a list is provided, multi-process encoding will be used. Defaults to None. normalize_embeddings (bool, optional): Whether to normalize returned vectors to have length 1. In that case, the faster dot-product (util.dot_score) instead of cosine similarity can be used. Defaults to False. truncate_dim (int, optional): The dimension to truncate sentence embeddings to. Truncation is especially interesting for `Matryoshka models `_, i.e. models that are trained to still produce useful embeddings even if the embedding dimension is reduced. Truncated embeddings require less memory and are faster to perform retrieval with, but note that inference is just as fast, and the embedding performance is worse than the full embeddings. If None, the ``truncate_dim`` from the model initialization is used. Defaults to None. pool (Dict[Literal["input", "output", "processes"], Any], optional): A pool created by `start_multi_process_pool()` for multi-process encoding. If provided, the encoding will be distributed across multiple processes. This is recommended for large datasets and when multiple GPUs are available. Defaults to None. chunk_size (int, optional): Size of chunks for multi-process encoding. Only used with multiprocessing, i.e. when ``pool`` is not None or ``device`` is a list. If None, a sensible default is calculated. Defaults to None. Returns: Union[List[Tensor], ndarray, Tensor]: By default, a 2d numpy array with shape [num_inputs, output_dimension] is returned. If only one string input is provided, then the output is a 1d array with shape [output_dimension]. If ``convert_to_tensor``, a torch Tensor is returned instead. If ``self.truncate_dim <= output_dimension`` then output_dimension is ``self.truncate_dim``. Example: :: from sentence_transformers import SentenceTransformer # Load a pre-trained SentenceTransformer model model = SentenceTransformer("mixedbread-ai/mxbai-embed-large-v1") # Encode some documents documents = [ "This research paper discusses the effects of climate change on marine life.", "The article explores the history of artificial intelligence development.", "This document contains technical specifications for the new product line.", ] # Using document-specific encoding embeddings = model.encode_document(documents) print(embeddings.shape) # (3, 768) """ if prompt_name is None and prompt is None: for candidate_prompt_name in ["document", "passage", "corpus"]: if candidate_prompt_name in self.prompts: prompt_name = candidate_prompt_name break return self.encode( sentences=sentences, prompt_name=prompt_name, prompt=prompt, batch_size=batch_size, show_progress_bar=show_progress_bar, output_value=output_value, precision=precision, convert_to_numpy=convert_to_numpy, convert_to_tensor=convert_to_tensor, device=device, normalize_embeddings=normalize_embeddings, truncate_dim=truncate_dim, pool=pool, chunk_size=chunk_size, task="document", **kwargs, ) # Return a single tensor because we're passing a single sentence. @overload def encode( self, sentences: str, prompt_name: str | None = ..., prompt: str | None = ..., batch_size: int = ..., show_progress_bar: bool | None = ..., output_value: Literal["sentence_embedding", "token_embeddings"] = ..., precision: Literal["float32", "int8", "uint8", "binary", "ubinary"] = ..., convert_to_numpy: Literal[False] = ..., convert_to_tensor: bool = ..., device: str | list[str | torch.device] | None = ..., normalize_embeddings: bool = ..., truncate_dim: int | None = ..., pool: dict[Literal["input", "output", "processes"], Any] | None = ..., chunk_size: int | None = ..., **kwargs, ) -> Tensor: ... # Return a single array, because convert_to_numpy is True # and "sentence_embeddings" is passed @overload def encode( self, sentences: str | list[str] | np.ndarray, prompt_name: str | None = ..., prompt: str | None = ..., batch_size: int = ..., show_progress_bar: bool | None = ..., output_value: Literal["sentence_embedding"] = ..., precision: Literal["float32", "int8", "uint8", "binary", "ubinary"] = ..., convert_to_numpy: Literal[True] = ..., convert_to_tensor: Literal[False] = ..., device: str | list[str | torch.device] | None = ..., normalize_embeddings: bool = ..., truncate_dim: int | None = ..., pool: dict[Literal["input", "output", "processes"], Any] | None = ..., chunk_size: int | None = ..., **kwargs, ) -> np.ndarray: ... # Return a single tensor, because convert_to_tensor is True # and "sentence_embeddings" is passed @overload def encode( self, sentences: str | list[str] | np.ndarray, prompt_name: str | None = ..., prompt: str | None = ..., batch_size: int = ..., show_progress_bar: bool | None = ..., output_value: Literal["sentence_embedding"] = ..., precision: Literal["float32", "int8", "uint8", "binary", "ubinary"] = ..., convert_to_numpy: bool = ..., convert_to_tensor: Literal[True] = ..., device: str | list[str | torch.device] | None = ..., normalize_embeddings: bool = ..., truncate_dim: int | None = ..., pool: dict[Literal["input", "output", "processes"], Any] | None = ..., chunk_size: int | None = ..., **kwargs, ) -> Tensor: ... # Return a list of tensors. Value of convert_ doesn't matter. @overload def encode( self, sentences: list[str] | np.ndarray, prompt_name: str | None = ..., prompt: str | None = ..., batch_size: int = ..., show_progress_bar: bool | None = ..., output_value: Literal["sentence_embedding", "token_embeddings"] = ..., precision: Literal["float32", "int8", "uint8", "binary", "ubinary"] = ..., convert_to_numpy: bool = ..., convert_to_tensor: bool = ..., device: str | list[str | torch.device] | None = ..., normalize_embeddings: bool = ..., truncate_dim: int | None = ..., pool: dict[Literal["input", "output", "processes"], Any] | None = ..., chunk_size: int | None = ..., **kwargs, ) -> list[Tensor]: ... # Return a list of dict of features, ignore the conversion args. @overload def encode( self, sentences: list[str] | np.ndarray, prompt_name: str | None = ..., prompt: str | None = ..., batch_size: int = ..., show_progress_bar: bool | None = ..., output_value: None = ..., precision: Literal["float32", "int8", "uint8", "binary", "ubinary"] = ..., convert_to_numpy: bool = ..., convert_to_tensor: bool = ..., device: str | list[str | torch.device] | None = ..., normalize_embeddings: bool = ..., truncate_dim: int | None = ..., pool: dict[Literal["input", "output", "processes"], Any] | None = ..., chunk_size: int | None = ..., **kwargs, ) -> list[dict[str, Tensor]]: ... # Return a dict of features, ignore the conversion args. @overload def encode( self, sentences: str, prompt_name: str | None = ..., prompt: str | None = ..., batch_size: int = ..., show_progress_bar: bool | None = ..., output_value: None = ..., precision: Literal["float32", "int8", "uint8", "binary", "ubinary"] = ..., convert_to_numpy: bool = ..., convert_to_tensor: bool = ..., device: str | list[str | torch.device] | None = ..., normalize_embeddings: bool = ..., truncate_dim: int | None = ..., pool: dict[Literal["input", "output", "processes"], Any] | None = ..., chunk_size: int | None = ..., **kwargs, ) -> dict[str, Tensor]: ... # If "token_embeddings" is True, then the output is a single tensor. @overload def encode( self, sentences: str, prompt_name: str | None = ..., prompt: str | None = ..., batch_size: int = ..., show_progress_bar: bool | None = ..., output_value: Literal["token_embeddings"] = ..., precision: Literal["float32", "int8", "uint8", "binary", "ubinary"] = ..., convert_to_numpy: bool = ..., convert_to_tensor: bool = ..., device: str | list[str | torch.device] | None = ..., normalize_embeddings: bool = ..., truncate_dim: int | None = ..., pool: dict[Literal["input", "output", "processes"], Any] | None = ..., chunk_size: int | None = ..., **kwargs, ) -> Tensor: ... @torch.inference_mode() def encode( self, sentences: str | list[str] | np.ndarray, prompt_name: str | None = None, prompt: str | None = None, batch_size: int = 32, show_progress_bar: bool | None = None, output_value: Literal["sentence_embedding", "token_embeddings"] | None = "sentence_embedding", precision: Literal["float32", "int8", "uint8", "binary", "ubinary"] = "float32", convert_to_numpy: bool = True, convert_to_tensor: bool = False, device: str | list[str | torch.device] | None = None, normalize_embeddings: bool = False, truncate_dim: int | None = None, pool: dict[Literal["input", "output", "processes"], Any] | None = None, chunk_size: int | None = None, **kwargs, ) -> list[Tensor] | np.ndarray | Tensor | dict[str, Tensor] | list[dict[str, Tensor]]: """ Computes sentence embeddings. .. tip:: If you are unsure whether you should use :meth:`encode`, :meth:`encode_query`, or :meth:`encode_document`, your best bet is to use :meth:`encode_query` and :meth:`encode_document` for Information Retrieval tasks with clear query and document/passage distinction, and use :meth:`encode` for all other tasks. Note that :meth:`encode` is the most general method and can be used for any task, including Information Retrieval, and that if the model was not trained with predefined prompts and/or task types, then all three methods will return identical embeddings. Args: sentences (Union[str, List[str]]): The sentences to embed. prompt_name (Optional[str], optional): The name of the prompt to use for encoding. Must be a key in the `prompts` dictionary, which is either set in the constructor or loaded from the model configuration. For example if ``prompt_name`` is "query" and the ``prompts`` is {"query": "query: ", ...}, then the sentence "What is the capital of France?" will be encoded as "query: What is the capital of France?" because the sentence is appended to the prompt. If ``prompt`` is also set, this argument is ignored. Defaults to None. prompt (Optional[str], optional): The prompt to use for encoding. For example, if the prompt is "query: ", then the sentence "What is the capital of France?" will be encoded as "query: What is the capital of France?" because the sentence is appended to the prompt. If ``prompt`` is set, ``prompt_name`` is ignored. Defaults to None. batch_size (int, optional): The batch size used for the computation. Defaults to 32. show_progress_bar (bool, optional): Whether to output a progress bar when encode sentences. Defaults to None. output_value (Optional[Literal["sentence_embedding", "token_embeddings"]], optional): The type of embeddings to return: "sentence_embedding" to get sentence embeddings, "token_embeddings" to get wordpiece token embeddings, and `None`, to get all output values. Defaults to "sentence_embedding". precision (Literal["float32", "int8", "uint8", "binary", "ubinary"], optional): The precision to use for the embeddings. Can be "float32", "int8", "uint8", "binary", or "ubinary". All non-float32 precisions are quantized embeddings. Quantized embeddings are smaller in size and faster to compute, but may have a lower accuracy. They are useful for reducing the size of the embeddings of a corpus for semantic search, among other tasks. Defaults to "float32". convert_to_numpy (bool, optional): Whether the output should be a list of numpy vectors. If False, it is a list of PyTorch tensors. Defaults to True. convert_to_tensor (bool, optional): Whether the output should be one large tensor. Overwrites `convert_to_numpy`. Defaults to False. device (Union[str, List[str], None], optional): Device(s) to use for computation. Can be: - A single device string (e.g., "cuda:0", "cpu") for single-process encoding - A list of device strings (e.g., ["cuda:0", "cuda:1"], ["cpu", "cpu", "cpu", "cpu"]) to distribute encoding across multiple processes - None to auto-detect available device for single-process encoding If a list is provided, multi-process encoding will be used. Defaults to None. normalize_embeddings (bool, optional): Whether to normalize returned vectors to have length 1. In that case, the faster dot-product (util.dot_score) instead of cosine similarity can be used. Defaults to False. truncate_dim (int, optional): The dimension to truncate sentence embeddings to. Truncation is especially interesting for `Matryoshka models `_, i.e. models that are trained to still produce useful embeddings even if the embedding dimension is reduced. Truncated embeddings require less memory and are faster to perform retrieval with, but note that inference is just as fast, and the embedding performance is worse than the full embeddings. If None, the ``truncate_dim`` from the model initialization is used. Defaults to None. pool (Dict[Literal["input", "output", "processes"], Any], optional): A pool created by `start_multi_process_pool()` for multi-process encoding. If provided, the encoding will be distributed across multiple processes. This is recommended for large datasets and when multiple GPUs are available. Defaults to None. chunk_size (int, optional): Size of chunks for multi-process encoding. Only used with multiprocessing, i.e. when ``pool`` is not None or ``device`` is a list. If None, a sensible default is calculated. Defaults to None. Returns: Union[List[Tensor], ndarray, Tensor]: By default, a 2d numpy array with shape [num_inputs, output_dimension] is returned. If only one string input is provided, then the output is a 1d array with shape [output_dimension]. If ``convert_to_tensor``, a torch Tensor is returned instead. If ``self.truncate_dim <= output_dimension`` then output_dimension is ``self.truncate_dim``. Example: :: from sentence_transformers import SentenceTransformer # Load a pre-trained SentenceTransformer model model = SentenceTransformer("all-mpnet-base-v2") # Encode some texts sentences = [ "The weather is lovely today.", "It's so sunny outside!", "He drove to the stadium.", ] embeddings = model.encode(sentences) print(embeddings.shape) # (3, 768) """ if self.device.type == "hpu" and not self.is_hpu_graph_enabled: import habana_frameworks.torch as ht if hasattr(ht, "hpu") and hasattr(ht.hpu, "wrap_in_hpu_graph"): ht.hpu.wrap_in_hpu_graph(self, disable_tensor_cache=True) self.is_hpu_graph_enabled = True self.eval() if show_progress_bar is None: show_progress_bar = logger.getEffectiveLevel() in (logging.INFO, logging.DEBUG) if convert_to_tensor: convert_to_numpy = False if output_value != "sentence_embedding": convert_to_tensor = False convert_to_numpy = False # Cast an individual input to a list with length 1 input_was_string = False if isinstance(sentences, str) or not hasattr(sentences, "__len__"): sentences = [sentences] input_was_string = True # If pool or a list of devices is provided, use multi-process encoding if pool is not None or (isinstance(device, list) and len(device) > 0): return self._encode_multi_process( sentences, # Utility and post-processing parameters show_progress_bar=show_progress_bar, input_was_string=input_was_string, # Multi-process encoding parameters pool=pool, device=device, chunk_size=chunk_size, # Encoding parameters prompt_name=prompt_name, prompt=prompt, batch_size=batch_size, output_value=output_value, precision=precision, convert_to_numpy=convert_to_numpy, convert_to_tensor=convert_to_tensor, normalize_embeddings=normalize_embeddings, truncate_dim=truncate_dim, **kwargs, ) # Original encoding logic when not using multi-process allowed_precisions = {"float32", "int8", "uint8", "binary", "ubinary"} if precision and precision not in allowed_precisions: raise ValueError(f"Precision {precision!r} is not supported") if prompt is None: if prompt_name is not None: try: prompt = self.prompts[prompt_name] except KeyError: raise ValueError( f"Prompt name '{prompt_name}' not found in the configured prompts dictionary with keys {list(self.prompts.keys())!r}." ) elif self.default_prompt_name is not None: prompt = self.prompts.get(self.default_prompt_name, None) else: if prompt_name is not None: logger.warning( "Encode with either a `prompt`, a `prompt_name`, or neither, but not both. " "Ignoring the `prompt_name` in favor of `prompt`." ) extra_features = {} if prompt is not None and len(prompt) > 0: sentences = [prompt + sentence for sentence in sentences] # Some models (e.g. INSTRUCTOR, GRIT) require removing the prompt before pooling # Tracking the prompt length allow us to remove the prompt during pooling length = self._get_prompt_length(prompt, **kwargs) if length is not None: extra_features["prompt_length"] = length # Here, device is either a single device string (e.g., "cuda:0", "cpu") for single-process encoding or None if device is None: device = self.device self.to(device) truncate_dim = truncate_dim if truncate_dim is not None else self.truncate_dim all_embeddings = [] length_sorted_idx = np.argsort([-self._text_length(sen) for sen in sentences]) sentences_sorted = [sentences[int(idx)] for idx in length_sorted_idx] for start_index in trange(0, len(sentences), batch_size, desc="Batches", disable=not show_progress_bar): sentences_batch = sentences_sorted[start_index : start_index + batch_size] features = self.tokenize(sentences_batch, **kwargs) if self.device.type == "hpu": if "input_ids" in features: curr_tokenize_len = features["input_ids"].shape additional_pad_len = 2 ** math.ceil(math.log2(curr_tokenize_len[1])) - curr_tokenize_len[1] features["input_ids"] = torch.cat( ( features["input_ids"], torch.ones((curr_tokenize_len[0], additional_pad_len), dtype=torch.int8), ), -1, ) features["attention_mask"] = torch.cat( ( features["attention_mask"], torch.zeros((curr_tokenize_len[0], additional_pad_len), dtype=torch.int8), ), -1, ) if "token_type_ids" in features: features["token_type_ids"] = torch.cat( ( features["token_type_ids"], torch.zeros((curr_tokenize_len[0], additional_pad_len), dtype=torch.int8), ), -1, ) features = batch_to_device(features, device) features.update(extra_features) with torch.no_grad(): out_features = self.forward(features, **kwargs) if self.device.type == "hpu": out_features = copy.deepcopy(out_features) if truncate_dim: out_features["sentence_embedding"] = truncate_embeddings( out_features["sentence_embedding"], truncate_dim ) if output_value == "token_embeddings": embeddings = [] for token_emb, attention in zip(out_features[output_value], out_features["attention_mask"]): last_mask_id = len(attention) - 1 while last_mask_id > 0 and attention[last_mask_id].item() == 0: last_mask_id -= 1 embeddings.append(token_emb[0 : last_mask_id + 1]) elif output_value is None: # Return all outputs embeddings = [] for idx in range(len(out_features["sentence_embedding"])): batch_item = {} for name, value in out_features.items(): try: batch_item[name] = value[idx] except TypeError: # Handle non-indexable values (like prompt_length) batch_item[name] = value embeddings.append(batch_item) else: # Sentence embeddings embeddings = out_features[output_value] embeddings = embeddings.detach() if normalize_embeddings: embeddings = torch.nn.functional.normalize(embeddings, p=2, dim=1) # fixes for #522 and #487 to avoid oom problems on gpu with large datasets if convert_to_numpy: embeddings = embeddings.cpu() all_embeddings.extend(embeddings) all_embeddings = [all_embeddings[idx] for idx in np.argsort(length_sorted_idx)] if all_embeddings and precision and precision != "float32": all_embeddings = quantize_embeddings(all_embeddings, precision=precision) if convert_to_tensor: if len(all_embeddings): if isinstance(all_embeddings, np.ndarray): all_embeddings = torch.from_numpy(all_embeddings) else: all_embeddings = torch.stack(all_embeddings) else: all_embeddings = torch.tensor([], device=self.device) elif convert_to_numpy: if not isinstance(all_embeddings, np.ndarray): if all_embeddings and all_embeddings[0].dtype == torch.bfloat16: all_embeddings = np.asarray([emb.float().numpy() for emb in all_embeddings]) else: all_embeddings = np.asarray([emb.numpy() for emb in all_embeddings]) elif isinstance(all_embeddings, np.ndarray): all_embeddings = [torch.from_numpy(embedding) for embedding in all_embeddings] if input_was_string: all_embeddings = all_embeddings[0] return all_embeddings def forward(self, input: dict[str, Tensor], **kwargs) -> dict[str, Tensor]: for module_name, module in self.named_children(): module_kwargs = {} if isinstance(module, Router): module_kwargs = kwargs else: module_kwarg_keys = [] if self.module_kwargs is not None: module_kwarg_keys = self.module_kwargs.get(module_name, []) module_kwargs = { key: value for key, value in kwargs.items() if key in module_kwarg_keys or (hasattr(module, "forward_kwargs") and key in module.forward_kwargs) } input = module(input, **module_kwargs) return input @property def similarity_fn_name(self) -> Literal["cosine", "dot", "euclidean", "manhattan"]: """Return the name of the similarity function used by :meth:`SentenceTransformer.similarity` and :meth:`SentenceTransformer.similarity_pairwise`. Returns: Optional[str]: The name of the similarity function. Can be None if not set, in which case it will default to "cosine" when first called. Example: >>> model = SentenceTransformer("multi-qa-mpnet-base-dot-v1") >>> model.similarity_fn_name 'dot' """ if self._similarity_fn_name is None: self.similarity_fn_name = SimilarityFunction.COSINE return self._similarity_fn_name @similarity_fn_name.setter def similarity_fn_name( self, value: Literal["cosine", "dot", "euclidean", "manhattan"] | SimilarityFunction ) -> None: if isinstance(value, SimilarityFunction): value = value.value self._similarity_fn_name = value if value is not None: self._similarity = SimilarityFunction.to_similarity_fn(value) self._similarity_pairwise = SimilarityFunction.to_similarity_pairwise_fn(value) @overload def similarity(self, embeddings1: Tensor, embeddings2: Tensor) -> Tensor: ... @overload def similarity(self, embeddings1: npt.NDArray[np.float32], embeddings2: npt.NDArray[np.float32]) -> Tensor: ... @property def similarity(self) -> Callable[[Tensor | npt.NDArray[np.float32], Tensor | npt.NDArray[np.float32]], Tensor]: """ Compute the similarity between two collections of embeddings. The output will be a matrix with the similarity scores between all embeddings from the first parameter and all embeddings from the second parameter. This differs from `similarity_pairwise` which computes the similarity between each pair of embeddings. This method supports only embeddings with fp32 precision and does not accommodate quantized embeddings. Args: embeddings1 (Union[Tensor, ndarray]): [num_embeddings_1, embedding_dim] or [embedding_dim]-shaped numpy array or torch tensor. embeddings2 (Union[Tensor, ndarray]): [num_embeddings_2, embedding_dim] or [embedding_dim]-shaped numpy array or torch tensor. Returns: Tensor: A [num_embeddings_1, num_embeddings_2]-shaped torch tensor with similarity scores. Example: :: >>> model = SentenceTransformer("all-mpnet-base-v2") >>> sentences = [ ... "The weather is so nice!", ... "It's so sunny outside.", ... "He's driving to the movie theater.", ... "She's going to the cinema.", ... ] >>> embeddings = model.encode(sentences, normalize_embeddings=True) >>> model.similarity(embeddings, embeddings) tensor([[1.0000, 0.7235, 0.0290, 0.1309], [0.7235, 1.0000, 0.0613, 0.1129], [0.0290, 0.0613, 1.0000, 0.5027], [0.1309, 0.1129, 0.5027, 1.0000]]) >>> model.similarity_fn_name "cosine" >>> model.similarity_fn_name = "euclidean" >>> model.similarity(embeddings, embeddings) tensor([[-0.0000, -0.7437, -1.3935, -1.3184], [-0.7437, -0.0000, -1.3702, -1.3320], [-1.3935, -1.3702, -0.0000, -0.9973], [-1.3184, -1.3320, -0.9973, -0.0000]]) """ if self.similarity_fn_name is None: self.similarity_fn_name = SimilarityFunction.COSINE return self._similarity @overload def similarity_pairwise(self, embeddings1: Tensor, embeddings2: Tensor) -> Tensor: ... @overload def similarity_pairwise( self, embeddings1: npt.NDArray[np.float32], embeddings2: npt.NDArray[np.float32] ) -> Tensor: ... @property def similarity_pairwise( self, ) -> Callable[[Tensor | npt.NDArray[np.float32], Tensor | npt.NDArray[np.float32]], Tensor]: """ Compute the similarity between two collections of embeddings. The output will be a vector with the similarity scores between each pair of embeddings. This method supports only embeddings with fp32 precision and does not accommodate quantized embeddings. Args: embeddings1 (Union[Tensor, ndarray]): [num_embeddings, embedding_dim] or [embedding_dim]-shaped numpy array or torch tensor. embeddings2 (Union[Tensor, ndarray]): [num_embeddings, embedding_dim] or [embedding_dim]-shaped numpy array or torch tensor. Returns: Tensor: A [num_embeddings]-shaped torch tensor with pairwise similarity scores. Example: :: >>> model = SentenceTransformer("all-mpnet-base-v2") >>> sentences = [ ... "The weather is so nice!", ... "It's so sunny outside.", ... "He's driving to the movie theater.", ... "She's going to the cinema.", ... ] >>> embeddings = model.encode(sentences, normalize_embeddings=True) >>> model.similarity_pairwise(embeddings[::2], embeddings[1::2]) tensor([0.7235, 0.5027]) >>> model.similarity_fn_name "cosine" >>> model.similarity_fn_name = "euclidean" >>> model.similarity_pairwise(embeddings[::2], embeddings[1::2]) tensor([-0.7437, -0.9973]) """ if self.similarity_fn_name is None: self.similarity_fn_name = SimilarityFunction.COSINE return self._similarity_pairwise def start_multi_process_pool( self, target_devices: list[str] | None = None ) -> dict[Literal["input", "output", "processes"], Any]: """ Starts a multi-process pool to process the encoding with several independent processes via :meth:`SentenceTransformer.encode_multi_process `. This method is recommended if you want to encode on multiple GPUs or CPUs. It is advised to start only one process per GPU. This method works together with encode_multi_process and stop_multi_process_pool. Args: target_devices (List[str], optional): PyTorch target devices, e.g. ["cuda:0", "cuda:1", ...], ["npu:0", "npu:1", ...], or ["cpu", "cpu", "cpu", "cpu"]. If target_devices is None and CUDA/NPU is available, then all available CUDA/NPU devices will be used. If target_devices is None and CUDA/NPU is not available, then 4 CPU devices will be used. Returns: Dict[str, Any]: A dictionary with the target processes, an input queue, and an output queue. """ if target_devices is None: if torch.cuda.is_available(): target_devices = [f"cuda:{i}" for i in range(torch.cuda.device_count())] elif is_torch_npu_available(): target_devices = [f"npu:{i}" for i in range(torch.npu.device_count())] else: logger.info("CUDA/NPU is not available. Starting 4 CPU workers") target_devices = ["cpu"] * 4 logger.info("Start multi-process pool on devices: {}".format(", ".join(map(str, target_devices)))) self.to("cpu") self.share_memory() ctx = mp.get_context("spawn") input_queue = ctx.Queue() output_queue = ctx.Queue() processes = [] for device_id in target_devices: p = ctx.Process( target=self.__class__._encode_multi_process_worker, args=(device_id, self, input_queue, output_queue), daemon=True, ) p.start() processes.append(p) return {"input": input_queue, "output": output_queue, "processes": processes} @staticmethod def stop_multi_process_pool(pool: dict[Literal["input", "output", "processes"], Any]) -> None: """ Stops all processes started with start_multi_process_pool. Args: pool (Dict[str, object]): A dictionary containing the input queue, output queue, and process list. Returns: None """ for p in pool["processes"]: p.terminate() for p in pool["processes"]: p.join() p.close() pool["input"].close() pool["output"].close() @deprecated( "The `encode_multi_process` method has been deprecated, and its functionality has been integrated into `encode`. " "You can now call `encode` with the same parameters to achieve multi-process encoding.", ) def encode_multi_process( self, sentences: list[str], pool: dict[Literal["input", "output", "processes"], Any], prompt_name: str | None = None, prompt: str | None = None, batch_size: int = 32, chunk_size: int | None = None, show_progress_bar: bool | None = None, precision: Literal["float32", "int8", "uint8", "binary", "ubinary"] = "float32", normalize_embeddings: bool = False, truncate_dim: int | None = None, ) -> np.ndarray: """ .. warning:: This method is deprecated. You can now call :meth:`SentenceTransformer.encode ` with the same parameters instead, which will automatically handle multi-process encoding using the provided ``pool``. Encodes a list of sentences using multiple processes and GPUs via :meth:`SentenceTransformer.encode `. The sentences are chunked into smaller packages and sent to individual processes, which encode them on different GPUs or CPUs. This method is only suitable for encoding large sets of sentences. Args: sentences (List[str]): List of sentences to encode. pool (Dict[Literal["input", "output", "processes"], Any]): A pool of workers started with :meth:`SentenceTransformer.start_multi_process_pool `. prompt_name (Optional[str], optional): The name of the prompt to use for encoding. Must be a key in the `prompts` dictionary, which is either set in the constructor or loaded from the model configuration. For example if ``prompt_name`` is "query" and the ``prompts`` is {"query": "query: ", ...}, then the sentence "What is the capital of France?" will be encoded as "query: What is the capital of France?" because the sentence is appended to the prompt. If ``prompt`` is also set, this argument is ignored. Defaults to None. prompt (Optional[str], optional): The prompt to use for encoding. For example, if the prompt is "query: ", then the sentence "What is the capital of France?" will be encoded as "query: What is the capital of France?" because the sentence is appended to the prompt. If ``prompt`` is set, ``prompt_name`` is ignored. Defaults to None. batch_size (int): Encode sentences with batch size. (default: 32) chunk_size (int): Sentences are chunked and sent to the individual processes. If None, it determines a sensible size. Defaults to None. show_progress_bar (bool, optional): Whether to output a progress bar when encode sentences. Defaults to None. precision (Literal["float32", "int8", "uint8", "binary", "ubinary"]): The precision to use for the embeddings. Can be "float32", "int8", "uint8", "binary", or "ubinary". All non-float32 precisions are quantized embeddings. Quantized embeddings are smaller in size and faster to compute, but may have lower accuracy. They are useful for reducing the size of the embeddings of a corpus for semantic search, among other tasks. Defaults to "float32". normalize_embeddings (bool): Whether to normalize returned vectors to have length 1. In that case, the faster dot-product (util.dot_score) instead of cosine similarity can be used. Defaults to False. truncate_dim (int, optional): The dimension to truncate sentence embeddings to. Truncation is especially interesting for `Matryoshka models `_, i.e. models that are trained to still produce useful embeddings even if the embedding dimension is reduced. Truncated embeddings require less memory and are faster to perform retrieval with, but note that inference is just as fast, and the embedding performance is worse than the full embeddings. If None, the ``truncate_dim`` from the model initialization is used. Defaults to None. Returns: np.ndarray: A 2D numpy array with shape [num_inputs, output_dimension]. Example: :: from sentence_transformers import SentenceTransformer def main(): model = SentenceTransformer("all-mpnet-base-v2") sentences = ["The weather is so nice!", "It's so sunny outside.", "He's driving to the movie theater.", "She's going to the cinema."] * 1000 pool = model.start_multi_process_pool() embeddings = model.encode_multi_process(sentences, pool) model.stop_multi_process_pool(pool) print(embeddings.shape) # => (4000, 768) if __name__ == "__main__": main() """ return self.encode( sentences, prompt_name=prompt_name, prompt=prompt, batch_size=batch_size, show_progress_bar=show_progress_bar, output_value="sentence_embedding", precision=precision, convert_to_numpy=True, convert_to_tensor=False, normalize_embeddings=normalize_embeddings, truncate_dim=truncate_dim, pool=pool, chunk_size=chunk_size, ) def _encode_multi_process( self, inputs: list[str], show_progress_bar: bool | None = True, input_was_string: bool = False, pool: dict[Literal["input", "output", "processes"], Any] | None = None, device: str | list[str | torch.device] | None = None, chunk_size: int | None = None, **encode_kwargs, ): convert_to_tensor = encode_kwargs.get("convert_to_tensor", False) convert_to_numpy = encode_kwargs.get("convert_to_numpy", False) encode_kwargs["show_progress_bar"] = False # Create a pool if is not provided, but a list of devices is created_pool = False if pool is None and isinstance(device, list): pool = self.start_multi_process_pool(device) created_pool = True try: # Determine chunk size if not provided. As a default, aim for 10 chunks per process, with a maximum of 5000 sentences per chunk. if chunk_size is None: chunk_size = min(math.ceil(len(inputs) / len(pool["processes"]) / 10), 5000) chunk_size = max(chunk_size, 1) # Ensure chunk_size is at least 1 input_queue: torch.multiprocessing.Queue = pool["input"] output_queue: torch.multiprocessing.Queue = pool["output"] # Send inputs to the input queue in chunks chunk_id = -1 # We default to -1 to handle empty input gracefully for chunk_id, chunk_start in enumerate(range(0, len(inputs), chunk_size)): chunk = inputs[chunk_start : chunk_start + chunk_size] input_queue.put([chunk_id, chunk, encode_kwargs]) # Collect results from the output queue output_list = sorted( [output_queue.get() for _ in trange(chunk_id + 1, desc="Chunks", disable=not show_progress_bar)], key=lambda x: x[0], ) if input_was_string: # If input was a single string, return the first (only) result directly return output_list[0][1][0] # Handle the various output formats: torch tensors, numpy arrays, or # list of dictionaries, also when empty. embeddings = [output[1] for output in output_list] if embeddings: if isinstance(embeddings[0], list): embeddings = sum(embeddings, []) elif isinstance(embeddings[0], torch.Tensor): embeddings = torch.cat(embeddings) elif isinstance(embeddings[0], np.ndarray): embeddings = np.concatenate(embeddings, axis=0) elif convert_to_tensor: embeddings = torch.Tensor() elif convert_to_numpy: embeddings = np.array([]) return embeddings finally: # Clean up the pool if we created it if created_pool: self.stop_multi_process_pool(pool) @staticmethod def _encode_multi_process_worker( target_device: str, model: SentenceTransformer, input_queue: Queue, results_queue: Queue ) -> None: """ Internal working process to encode sentences in multi-process setup """ while True: try: chunk_id, inputs, kwargs = input_queue.get() embeddings = model.encode(inputs, device=target_device, **kwargs) results_queue.put([chunk_id, embeddings]) except queue.Empty: break def set_pooling_include_prompt(self, include_prompt: bool) -> None: """ Sets the `include_prompt` attribute in the pooling layer in the model, if there is one. This is useful for INSTRUCTOR models, as the prompt should be excluded from the pooling strategy for these models. Args: include_prompt (bool): Whether to include the prompt in the pooling layer. Returns: None """ for module in self: if isinstance(module, Pooling): module.include_prompt = include_prompt break def _get_prompt_length(self, prompt: str, **kwargs) -> int: """ Return the length of the prompt in tokens, including the BOS token """ if (prompt, *kwargs.values()) in self._prompt_length_mapping: return self._prompt_length_mapping[(prompt, *kwargs.values())] tokenized_prompt = self.tokenize([prompt], **kwargs) if "input_ids" not in tokenized_prompt: # If the tokenizer does not return input_ids, we cannot determine the prompt length. # This can happen with some tokenizers that do not use input_ids. return None prompt_length = tokenized_prompt["input_ids"].shape[-1] # If the tokenizer adds a special EOS token, we do not count it as part of the prompt length. # This is to ensure that the prompt length does not include the EOS token. last_token = tokenized_prompt["input_ids"][..., -1].item() if hasattr(self.tokenizer, "all_special_ids") and last_token in self.tokenizer.all_special_ids: prompt_length -= 1 self._prompt_length_mapping[(prompt, *kwargs.values())] = prompt_length return prompt_length def get_max_seq_length(self) -> int | None: """ Returns the maximal sequence length that the model accepts. Longer inputs will be truncated. Returns: Optional[int]: The maximal sequence length that the model accepts, or None if it is not defined. """ if hasattr(self._first_module(), "max_seq_length"): return self._first_module().max_seq_length return None def tokenize(self, texts: list[str] | list[dict] | list[tuple[str, str]], **kwargs) -> dict[str, Tensor]: """ Tokenizes the texts. Args: texts (Union[List[str], List[Dict], List[Tuple[str, str]]]): A list of texts to be tokenized. Returns: Dict[str, Tensor]: A dictionary of tensors with the tokenized texts. Common keys are "input_ids", "attention_mask", and "token_type_ids". """ try: return self[0].tokenize(texts, **kwargs) except TypeError: return self[0].tokenize(texts) def get_sentence_features(self, *features) -> dict[Literal["sentence_embedding"], Tensor]: return self._first_module().get_sentence_features(*features) def get_sentence_embedding_dimension(self) -> int | None: """ Returns the number of dimensions in the output of :meth:`SentenceTransformer.encode `. Returns: Optional[int]: The number of dimensions in the output of `encode`. If it's not known, it's `None`. """ output_dim = None for mod in reversed(self._modules.values()): sent_embedding_dim_method = getattr(mod, "get_sentence_embedding_dimension", None) if callable(sent_embedding_dim_method): output_dim = sent_embedding_dim_method() break if self.truncate_dim is not None: # The user requested truncation. If they set it to a dim greater than output_dim, # no truncation will actually happen. So return output_dim instead of self.truncate_dim return min(output_dim or np.inf, self.truncate_dim) return output_dim @contextmanager def truncate_sentence_embeddings(self, truncate_dim: int | None) -> Iterator[None]: """ In this context, :meth:`SentenceTransformer.encode ` outputs sentence embeddings truncated at dimension ``truncate_dim``. This may be useful when you are using the same model for different applications where different dimensions are needed. Args: truncate_dim (int, optional): The dimension to truncate sentence embeddings to. ``None`` does no truncation. Example: :: from sentence_transformers import SentenceTransformer model = SentenceTransformer("all-mpnet-base-v2") with model.truncate_sentence_embeddings(truncate_dim=16): embeddings_truncated = model.encode(["hello there", "hiya"]) assert embeddings_truncated.shape[-1] == 16 """ original_output_dim = self.truncate_dim try: self.truncate_dim = truncate_dim yield finally: self.truncate_dim = original_output_dim def _first_module(self) -> torch.nn.Module: """Returns the first module of this sequential embedder""" return self._modules[next(iter(self._modules))] def _last_module(self) -> torch.nn.Module: """Returns the last module of this sequential embedder""" return self._modules[next(reversed(self._modules))] def save( self, path: str, model_name: str | None = None, create_model_card: bool = True, train_datasets: list[str] | None = None, safe_serialization: bool = True, ) -> None: """ Saves a model and its configuration files to a directory, so that it can be loaded with ``SentenceTransformer(path)`` again. Args: path (str): Path on disc where the model will be saved. model_name (str, optional): Optional model name. create_model_card (bool, optional): If True, create a README.md with basic information about this model. train_datasets (List[str], optional): Optional list with the names of the datasets used to train the model. safe_serialization (bool, optional): If True, save the model using safetensors. If False, save the model the traditional (but unsafe) PyTorch way. """ if path is None: return os.makedirs(path, exist_ok=True) logger.info(f"Save model to {path}") modules_config = [] # Save some model info self._model_config["__version__"] = { "sentence_transformers": __version__, "transformers": transformers.__version__, "pytorch": torch.__version__, } with open(os.path.join(path, "config_sentence_transformers.json"), "w", encoding="utf8") as fOut: config = self._model_config.copy() config["prompts"] = self.prompts config["default_prompt_name"] = self.default_prompt_name config["similarity_fn_name"] = self.similarity_fn_name json.dump(config, fOut, indent=2) # Save modules for idx, name in enumerate(self._modules): module: Module = self._modules[name] if ( idx == 0 and hasattr(module, "save_in_root") and module.save_in_root ): # Save first module in the main folder model_path = os.path.join(path, "") else: model_path = os.path.join(path, str(idx) + "_" + type(module).__name__) os.makedirs(model_path, exist_ok=True) # Try to save with safetensors, but fall back to the traditional PyTorch way if the module doesn't support it try: module.save(model_path, safe_serialization=safe_serialization) except TypeError: module.save(model_path) # "module" only works for Sentence Transformers as the modules have the same names as the classes class_ref = type(module).__module__ # For remote modules, we want to remove "transformers_modules.{repo_name}": if class_ref.startswith("transformers_modules."): class_file = sys.modules[class_ref].__file__ # Save the custom module file dest_file = Path(model_path) / (Path(class_file).name) shutil.copy(class_file, dest_file) # Save all files importeed in the custom module file for needed_file in get_relative_import_files(class_file): dest_file = Path(model_path) / (Path(needed_file).name) shutil.copy(needed_file, dest_file) # For remote modules, we want to ignore the "transformers_modules.{repo_id}" part, # i.e. we only want the filename class_ref = f"{class_ref.split('.')[-1]}.{type(module).__name__}" # For other cases, we want to add the class name: elif not class_ref.startswith("sentence_transformers."): class_ref = f"{class_ref}.{type(module).__name__}" module_config = {"idx": idx, "name": name, "path": os.path.basename(model_path), "type": class_ref} if self.module_kwargs and name in self.module_kwargs and (module_kwargs := self.module_kwargs[name]): module_config["kwargs"] = module_kwargs modules_config.append(module_config) with open(os.path.join(path, "modules.json"), "w", encoding="utf8") as fOut: json.dump(modules_config, fOut, indent=2) # Create model card if create_model_card: self._create_model_card(path, model_name, train_datasets) def save_pretrained( self, path: str, model_name: str | None = None, create_model_card: bool = True, train_datasets: list[str] | None = None, safe_serialization: bool = True, ) -> None: """ Saves a model and its configuration files to a directory, so that it can be loaded with ``SentenceTransformer(path)`` again. Args: path (str): Path on disc where the model will be saved. model_name (str, optional): Optional model name. create_model_card (bool, optional): If True, create a README.md with basic information about this model. train_datasets (List[str], optional): Optional list with the names of the datasets used to train the model. safe_serialization (bool, optional): If True, save the model using safetensors. If False, save the model the traditional (but unsafe) PyTorch way. """ self.save( path, model_name=model_name, create_model_card=create_model_card, train_datasets=train_datasets, safe_serialization=safe_serialization, ) def _update_default_model_id(self, model_card): if self.model_card_data.model_id: model_card = model_card.replace( 'model = SentenceTransformer("sentence_transformers_model_id"', f'model = SentenceTransformer("{self.model_card_data.model_id}"', ) return model_card def _create_model_card( self, path: str, model_name: str | None = None, train_datasets: list[str] | None = "deprecated" ) -> None: """ Create an automatic model and stores it in the specified path. If no training was done and the loaded model was a Sentence Transformer model already, then its model card is reused. Args: path (str): The path where the model card will be stored. model_name (Optional[str], optional): The name of the model. Defaults to None. train_datasets (Optional[List[str]], optional): Deprecated argument. Defaults to "deprecated". Returns: None """ if model_name: model_path = Path(model_name) if not model_path.exists() and not self.model_card_data.model_id: self.model_card_data.model_id = model_name # If we loaded a Sentence Transformer model from the Hub, and no training was done, then # we don't generate a new model card, but reuse the old one instead. if self._model_card_text and "generated_from_trainer" not in self.model_card_data.tags: model_card = self._model_card_text model_card = self._update_default_model_id(model_card) else: try: model_card = generate_model_card(self) except Exception: logger.error( f"Error while generating model card:\n{traceback.format_exc()}" "Consider opening an issue on https://github.com/UKPLab/sentence-transformers/issues with this traceback.\n" "Skipping model card creation." ) return with open(os.path.join(path, "README.md"), "w", encoding="utf8") as fOut: fOut.write(model_card) @save_to_hub_args_decorator def save_to_hub( self, repo_id: str, organization: str | None = None, token: str | None = None, private: bool | None = None, safe_serialization: bool = True, commit_message: str = "Add new SentenceTransformer model.", local_model_path: str | None = None, exist_ok: bool = False, replace_model_card: bool = False, train_datasets: list[str] | None = None, ) -> str: """ DEPRECATED, use `push_to_hub` instead. Uploads all elements of this Sentence Transformer to a new HuggingFace Hub repository. Args: repo_id (str): Repository name for your model in the Hub, including the user or organization. token (str, optional): An authentication token (See https://huggingface.co/settings/token) private (bool, optional): Set to true, for hosting a private model safe_serialization (bool, optional): If true, save the model using safetensors. If false, save the model the traditional PyTorch way commit_message (str, optional): Message to commit while pushing. local_model_path (str, optional): Path of the model locally. If set, this file path will be uploaded. Otherwise, the current model will be uploaded exist_ok (bool, optional): If true, saving to an existing repository is OK. If false, saving only to a new repository is possible replace_model_card (bool, optional): If true, replace an existing model card in the hub with the automatically created model card train_datasets (List[str], optional): Datasets used to train the model. If set, the datasets will be added to the model card in the Hub. Returns: str: The url of the commit of your model in the repository on the Hugging Face Hub. """ logger.warning( "The `save_to_hub` method is deprecated and will be removed in a future version of SentenceTransformers." " Please use `push_to_hub` instead for future model uploads." ) if organization: if "/" not in repo_id: logger.warning( f'Providing an `organization` to `save_to_hub` is deprecated, please use `repo_id="{organization}/{repo_id}"` instead.' ) repo_id = f"{organization}/{repo_id}" elif repo_id.split("/")[0] != organization: raise ValueError( "Providing an `organization` to `save_to_hub` is deprecated, please only use `repo_id`." ) else: logger.warning( f'Providing an `organization` to `save_to_hub` is deprecated, please only use `repo_id="{repo_id}"` instead.' ) return self.push_to_hub( repo_id=repo_id, token=token, private=private, safe_serialization=safe_serialization, commit_message=commit_message, local_model_path=local_model_path, exist_ok=exist_ok, replace_model_card=replace_model_card, train_datasets=train_datasets, ) def push_to_hub( self, repo_id: str, token: str | None = None, private: bool | None = None, safe_serialization: bool = True, commit_message: str | None = None, local_model_path: str | None = None, exist_ok: bool = False, replace_model_card: bool = False, train_datasets: list[str] | None = None, revision: str | None = None, create_pr: bool = False, ) -> str: """ Uploads all elements of this Sentence Transformer to a new HuggingFace Hub repository. Args: repo_id (str): Repository name for your model in the Hub, including the user or organization. token (str, optional): An authentication token (See https://huggingface.co/settings/token) private (bool, optional): Set to true, for hosting a private model safe_serialization (bool, optional): If true, save the model using safetensors. If false, save the model the traditional PyTorch way commit_message (str, optional): Message to commit while pushing. local_model_path (str, optional): Path of the model locally. If set, this file path will be uploaded. Otherwise, the current model will be uploaded exist_ok (bool, optional): If true, saving to an existing repository is OK. If false, saving only to a new repository is possible replace_model_card (bool, optional): If true, replace an existing model card in the hub with the automatically created model card train_datasets (List[str], optional): Datasets used to train the model. If set, the datasets will be added to the model card in the Hub. revision (str, optional): Branch to push the uploaded files to create_pr (bool, optional): If True, create a pull request instead of pushing directly to the main branch Returns: str: The url of the commit of your model in the repository on the Hugging Face Hub. """ api = HfApi(token=token) repo_url = api.create_repo( repo_id=repo_id, private=private, repo_type=None, exist_ok=exist_ok or create_pr, ) repo_id = repo_url.repo_id # Update the repo_id in case the old repo_id didn't contain a user or organization self.model_card_data.set_model_id(repo_id) if revision is not None: api.create_branch(repo_id=repo_id, branch=revision, exist_ok=True) if commit_message is None: backend = self.get_backend() if backend == "torch": commit_message = f"Add new {self.__class__.__name__} model" else: commit_message = f"Add new {self.__class__.__name__} model with an {backend} backend" commit_description = "" if create_pr: commit_description = f"""\ Hello! *This pull request has been automatically generated from the [`push_to_hub`](https://sbert.net/docs/package_reference/sentence_transformer/SentenceTransformer.html#sentence_transformers.SentenceTransformer.push_to_hub) method from the Sentence Transformers library.* ## Full Model Architecture: ``` {self} ``` ## Tip: Consider testing this pull request before merging by loading the model from this PR with the `revision` argument: ```python from sentence_transformers import {self.__class__.__name__} # TODO: Fill in the PR number pr_number = 2 model = {self.__class__.__name__}( "{repo_id}", revision=f"refs/pr/{{pr_number}}", backend="{self.get_backend()}", ) # Verify that everything works as expected embeddings = model.encode(["The weather is lovely today.", "It's so sunny outside!", "He drove to the stadium."]) print(embeddings.shape) similarities = model.similarity(embeddings, embeddings) print(similarities) ``` """ if local_model_path: folder_url = api.upload_folder( repo_id=repo_id, folder_path=local_model_path, commit_message=commit_message, commit_description=commit_description, revision=revision, create_pr=create_pr, ) else: with tempfile.TemporaryDirectory() as tmp_dir: create_model_card = replace_model_card or not os.path.exists(os.path.join(tmp_dir, "README.md")) self.save_pretrained( tmp_dir, model_name=repo_url.repo_id, create_model_card=create_model_card, train_datasets=train_datasets, safe_serialization=safe_serialization, ) folder_url = api.upload_folder( repo_id=repo_id, folder_path=tmp_dir, commit_message=commit_message, commit_description=commit_description, revision=revision, create_pr=create_pr, ) if create_pr: return folder_url.pr_url return folder_url.commit_url def _text_length(self, text: list[int] | list[list[int]]) -> int: """ Help function to get the length for the input text. Text can be either a list of ints (which means a single text as input), or a tuple of list of ints (representing several text inputs to the model). """ if isinstance(text, dict): # {key: value} case return len(next(iter(text.values()))) elif not hasattr(text, "__len__"): # Object has no len() method return 1 elif len(text) == 0 or isinstance(text[0], int): # Empty string or list of ints return len(text) else: return sum([len(t) for t in text]) # Sum of length of individual strings def evaluate(self, evaluator: SentenceEvaluator, output_path: str | None = None) -> dict[str, float] | float: """ Evaluate the model based on an evaluator Args: evaluator (SentenceEvaluator): The evaluator used to evaluate the model. output_path (str, optional): The path where the evaluator can write the results. Defaults to None. Returns: The evaluation results. """ if output_path is not None: os.makedirs(output_path, exist_ok=True) return evaluator(self, output_path) def _load_auto_model( self, model_name_or_path: str, token: bool | str | None, cache_folder: str | None, revision: str | None = None, trust_remote_code: bool = False, local_files_only: bool = False, model_kwargs: dict[str, Any] | None = None, tokenizer_kwargs: dict[str, Any] | None = None, config_kwargs: dict[str, Any] | None = None, has_modules: bool = False, ) -> list[nn.Module]: """ Creates a simple Transformer + Mean Pooling model and returns the modules Args: model_name_or_path (str): The name or path of the pre-trained model. token (Optional[Union[bool, str]]): The token to use for the model. cache_folder (Optional[str]): The folder to cache the model. revision (Optional[str], optional): The revision of the model. Defaults to None. trust_remote_code (bool, optional): Whether to trust remote code. Defaults to False. local_files_only (bool, optional): Whether to use only local files. Defaults to False. model_kwargs (Optional[Dict[str, Any]], optional): Additional keyword arguments for the model. Defaults to None. tokenizer_kwargs (Optional[Dict[str, Any]], optional): Additional keyword arguments for the tokenizer. Defaults to None. config_kwargs (Optional[Dict[str, Any]], optional): Additional keyword arguments for the config. Defaults to None. has_modules (bool, optional): Whether the model has modules.json. Defaults to False. Returns: List[nn.Module]: A list containing the transformer model and the pooling model. """ logger.warning( f"No sentence-transformers model found with name {model_name_or_path}. Creating a new one with mean pooling." ) shared_kwargs = { "token": token, "trust_remote_code": trust_remote_code, "revision": revision, "local_files_only": local_files_only, } model_kwargs = shared_kwargs if model_kwargs is None else {**shared_kwargs, **model_kwargs} tokenizer_kwargs = shared_kwargs if tokenizer_kwargs is None else {**shared_kwargs, **tokenizer_kwargs} config_kwargs = shared_kwargs if config_kwargs is None else {**shared_kwargs, **config_kwargs} transformer_model = Transformer( model_name_or_path, cache_dir=cache_folder, model_args=model_kwargs, tokenizer_args=tokenizer_kwargs, config_args=config_kwargs, backend=self.backend, ) pooling_model = Pooling(transformer_model.get_word_embedding_dimension(), "mean") if not local_files_only: self.model_card_data.set_base_model(model_name_or_path, revision=revision) return [transformer_model, pooling_model] def _load_module_class_from_ref( self, class_ref: str, model_name_or_path: str, trust_remote_code: bool, revision: str | None, model_kwargs: dict[str, Any] | None, ) -> nn.Module: # If the class is from sentence_transformers, we can directly import it, # otherwise, we try to import it dynamically, and if that fails, we fall back to the default import if class_ref.startswith("sentence_transformers."): return import_from_string(class_ref) if trust_remote_code or os.path.exists(model_name_or_path): code_revision = model_kwargs.pop("code_revision", None) if model_kwargs else None try: return get_class_from_dynamic_module( class_ref, model_name_or_path, revision=revision, code_revision=code_revision, ) except (OSError, ValueError): # Ignore the error if 1) the file does not exist, or 2) the class_ref is not correctly formatted/found pass return import_from_string(class_ref) def _load_sbert_model( self, model_name_or_path: str, token: bool | str | None, cache_folder: str | None, revision: str | None = None, trust_remote_code: bool = False, local_files_only: bool = False, model_kwargs: dict[str, Any] | None = None, tokenizer_kwargs: dict[str, Any] | None = None, config_kwargs: dict[str, Any] | None = None, ) -> dict[str, nn.Module]: """ Loads a full SentenceTransformer model using the modules.json file. Args: model_name_or_path (str): The name or path of the pre-trained model. token (Optional[Union[bool, str]]): The token to use for the model. cache_folder (Optional[str]): The folder to cache the model. revision (Optional[str], optional): The revision of the model. Defaults to None. trust_remote_code (bool, optional): Whether to trust remote code. Defaults to False. local_files_only (bool, optional): Whether to use only local files. Defaults to False. model_kwargs (Optional[Dict[str, Any]], optional): Additional keyword arguments for the model. Defaults to None. tokenizer_kwargs (Optional[Dict[str, Any]], optional): Additional keyword arguments for the tokenizer. Defaults to None. config_kwargs (Optional[Dict[str, Any]], optional): Additional keyword arguments for the config. Defaults to None. Returns: OrderedDict[str, nn.Module]: An ordered dictionary containing the modules of the model. """ # Check if the config_sentence_transformers.json file exists (exists since v2 of the framework) config_sentence_transformers_json_path = load_file_path( model_name_or_path, "config_sentence_transformers.json", token=token, cache_folder=cache_folder, revision=revision, local_files_only=local_files_only, ) if config_sentence_transformers_json_path is not None: with open(config_sentence_transformers_json_path, encoding="utf8") as fIn: self._model_config = json.load(fIn) if ( "__version__" in self._model_config and "sentence_transformers" in self._model_config["__version__"] and version.parse(self._model_config["__version__"]["sentence_transformers"]) > version.parse(__version__) ): logger.warning( f"You are trying to use a model that was created with Sentence Transformers version {self._model_config['__version__']['sentence_transformers']}, " f"but you're currently using version {__version__}. This might cause unexpected behavior or errors. " "In that case, try to update to the latest version." ) # Set score functions & prompts if not already overridden by the __init__ calls if self._similarity_fn_name is None: self.similarity_fn_name = self._model_config.get("similarity_fn_name", None) # Only update prompts that aren't already set by the user or defaults for prompt_name, prompt_text in self._model_config.get("prompts", {}).items(): if prompt_name not in self.prompts or not self.prompts[prompt_name]: self.prompts[prompt_name] = prompt_text if not self.default_prompt_name: self.default_prompt_name = self._model_config.get("default_prompt_name", None) if "model_type" not in self._model_config.keys(): self._model_config["model_type"] = self.__class__.__name__ # Check if a readme exists model_card_path = load_file_path( model_name_or_path, "README.md", token=token, cache_folder=cache_folder, revision=revision, local_files_only=local_files_only, ) if model_card_path is not None: try: with open(model_card_path, encoding="utf8") as fIn: self._model_card_text = fIn.read() except Exception: pass # Load the modules of sentence transformer modules_json_path = load_file_path( model_name_or_path, "modules.json", token=token, cache_folder=cache_folder, revision=revision, local_files_only=local_files_only, ) with open(modules_json_path, encoding="utf8") as fIn: modules_config = json.load(fIn) modules = OrderedDict() module_kwargs = OrderedDict() for module_config in modules_config: class_ref = module_config["type"] module_class: Module = self._load_module_class_from_ref( class_ref, model_name_or_path, trust_remote_code, revision, model_kwargs ) # Backwards compatibility: if the module is older and its `load` method only supports one parameter, # a path to a local directory containing the module files, then we load it with the old style load_signature = inspect.signature(module_class.load) # Check if the `load` method only accepts a single parameter (the path to the local directory). # This indicates an older module that does not support the newer loading method with multiple arguments. if len(load_signature.parameters) == 1: signature = inspect.signature(module_class.__init__) # If the module's `__init__` method contains specific keyword arguments like `model_args` and `config_args`, # it is likely Transformer-based. These arguments are commonly used in Transformer models to configure # the model and tokenizer during initialization. # Example: Models with custom modules on the Hugging Face Hub like # https://huggingface.co/jinaai/jina-embeddings-v3 may use this logic. if {"model_args", "config_args"} <= set(signature.parameters): # Load initialization arguments specific to Transformer-based modules. This includes # arguments for loading the model, tokenizer, and configuration, as well as any # additional module-specific keyword arguments. common_transformer_init_kwargs = Transformer._load_init_kwargs( model_name_or_path, # Loading-specific keyword arguments subfolder=module_config["path"], token=token, cache_folder=cache_folder, revision=revision, local_files_only=local_files_only, # Module-specific keyword arguments trust_remote_code=trust_remote_code, model_kwargs=model_kwargs, tokenizer_kwargs=tokenizer_kwargs, config_kwargs=config_kwargs, backend=self.backend, ) module = module_class(model_name_or_path, **common_transformer_init_kwargs) else: # Old modules that don't support the new loading method and don't seem Transformer-based # are loaded by downloading the full directories and calling .load() with the old style # (i.e. only a path to the local directory) local_path = load_dir_path( model_name_or_path=model_name_or_path, subfolder=module_config["path"], token=token, cache_folder=cache_folder, revision=revision, local_files_only=local_files_only, ) module = module_class.load(local_path) else: # Newer modules that support the new loading method are loaded with the new style # i.e. with many keyword arguments that can optionally be used by the modules module = module_class.load( model_name_or_path, # Loading-specific keyword arguments subfolder=module_config["path"], token=token, cache_folder=cache_folder, revision=revision, local_files_only=local_files_only, # Module-specific keyword arguments trust_remote_code=trust_remote_code, model_kwargs=model_kwargs, tokenizer_kwargs=tokenizer_kwargs, config_kwargs=config_kwargs, backend=self.backend, ) modules[module_config["name"]] = module module_kwargs[module_config["name"]] = module_config.get("kwargs", []) if revision is None: path_parts = Path(modules_json_path) if len(path_parts.parts) >= 2: revision_path_part = Path(modules_json_path).parts[-2] if len(revision_path_part) == 40: revision = revision_path_part if not local_files_only: self.model_card_data.set_base_model(model_name_or_path, revision=revision) return modules, module_kwargs @staticmethod @deprecated("SentenceTransformer.load(...) is deprecated, use SentenceTransformer(...) instead.") def load(input_path) -> SentenceTransformer: return SentenceTransformer(input_path) @property def device(self) -> device: """ Get torch.device from module, assuming that the whole module has one device. In case there are no PyTorch parameters, fall back to CPU. """ if (transformers_model := self.transformers_model) is not None and hasattr(transformers_model, "device"): return transformers_model.device if len(self._modules) and hasattr(self[0], "auto_model") and hasattr(self[0].auto_model, "device"): return self[0].auto_model.device try: return next(self.parameters()).device except StopIteration: # For nn.DataParallel compatibility in PyTorch 1.5 def find_tensor_attributes(module: nn.Module) -> list[tuple[str, Tensor]]: tuples = [(k, v) for k, v in module.__dict__.items() if torch.is_tensor(v)] return tuples gen = self._named_members(get_members_fn=find_tensor_attributes) try: first_tuple = next(gen) return first_tuple[1].device except StopIteration: return torch.device("cpu") @property def tokenizer(self) -> Any: """ Property to get the tokenizer that is used by this model """ return self._first_module().tokenizer @tokenizer.setter def tokenizer(self, value) -> None: """ Property to set the tokenizer that should be used by this model """ self._first_module().tokenizer = value @property def max_seq_length(self) -> int: """ Returns the maximal input sequence length for the model. Longer inputs will be truncated. Returns: int: The maximal input sequence length. Example: :: from sentence_transformers import SentenceTransformer model = SentenceTransformer("all-mpnet-base-v2") print(model.max_seq_length) # => 384 """ return self._first_module().max_seq_length @max_seq_length.setter def max_seq_length(self, value) -> None: """ Property to set the maximal input sequence length for the model. Longer inputs will be truncated. """ self._first_module().max_seq_length = value @property def transformers_model(self) -> PreTrainedModel | None: """ Property to get the underlying transformers PreTrainedModel instance, if it exists. Note that it's possible for a model to have multiple underlying transformers models, but this property will return the first one it finds in the module hierarchy. Returns: PreTrainedModel or None: The underlying transformers model or None if not found. Example: :: from sentence_transformers import SentenceTransformer model = SentenceTransformer("all-mpnet-base-v2") # You can now access the underlying transformers model transformers_model = model.transformers_model print(type(transformers_model)) # => """ for module in self.modules(): if isinstance(module, PreTrainedModel): return module return None @property def _target_device(self) -> torch.device: logger.warning( "`SentenceTransformer._target_device` has been deprecated, please use `SentenceTransformer.device` instead.", ) return self.device @_target_device.setter def _target_device(self, device: int | str | torch.device | None = None) -> None: self.to(device) @property def dtype(self) -> torch.dtype | None: for child in self.modules(): if child is not self and hasattr(child, "dtype"): return child.dtype return None @property def _no_split_modules(self) -> list[str]: try: return self._first_module()._no_split_modules except AttributeError: return [] @property def _keys_to_ignore_on_save(self) -> list[str]: try: return self._first_module()._keys_to_ignore_on_save except AttributeError: return [] def gradient_checkpointing_enable(self, gradient_checkpointing_kwargs=None) -> None: # Propagate the gradient checkpointing to the transformer model for child in self.modules(): if isinstance(child, PreTrainedModel): return child.gradient_checkpointing_enable(gradient_checkpointing_kwargs) def _get_model_type( self, model_name_or_path: str, token: bool | str | None, cache_folder: str | None, revision: str | None = None, local_files_only: bool = False, ) -> str | None: """ Retrieves the model_type from the config_sentence_transformers.json file. This is used to determine the appropriate loading method: - SentenceTransformer: These models should be loaded with _load_sbert_model when used with SentenceTransformer class - SparseEncoder: These models should be loaded with _load_auto_model when used with SentenceTransformer class When a model type doesn't match the class being used to load it, we switch loading methods to ensure compatibility. Args: model_name_or_path (str): The name or path of the pre-trained model. token (Optional[Union[bool, str]]): The token to use for the model. cache_folder (Optional[str]): The folder to cache the model. revision (Optional[str], optional): The revision of the model. Defaults to None. local_files_only (bool, optional): Whether to use only local files. Defaults to False. Returns: Optional[str]: The model type (SentenceTransformer or SparseEncoder) if available, None otherwise. """ config_sentence_transformers_json_path = load_file_path( model_name_or_path, "config_sentence_transformers.json", token=token, cache_folder=cache_folder, revision=revision, local_files_only=local_files_only, ) if config_sentence_transformers_json_path is None: return "SentenceTransformer" with open(config_sentence_transformers_json_path, encoding="utf8") as fIn: config = json.load(fIn) return config.get("model_type", "SentenceTransformer") # Default to "SentenceTransformer" if not specified