fastembed-gpu 0.5.0__tar.gz → 0.6.0__tar.gz

{fastembed_gpu-0.5.0 → fastembed_gpu-0.6.0}/NOTICE

@@ -7,8 +7,16 @@ This distribution includes the following Jina AI models, each with its respectiv
     - License: cc-by-nc-4.0
 - jinaai/jina-reranker-v2-base-multilingual
     - License: cc-by-nc-4.0
+- jinaai/jina-embeddings-v3
+    - License: cc-by-nc-4.0
 
 These models are developed by Jina (https://jina.ai/) and are subject to Jina AI's licensing terms.
 
+This distribution includes the following Google models, each with its respective license:
+- vidore/colpali-v1.3
+    - License: gemma
+
+Gemma is provided under and subject to the Gemma Terms of Use found at ai.google.dev/gemma/terms
+
 Additional Notes:
 This project also includes third-party libraries with their respective licenses. Please refer to the documentation of each library for details regarding its usage and licensing terms.

{fastembed_gpu-0.5.0 → fastembed_gpu-0.6.0}/PKG-INFO

@@ -1,8 +1,7 @@
-Metadata-Version: 2.1
+Metadata-Version: 2.3
 Name: fastembed-gpu
-Version: 0.5.0
+Version: 0.6.0
 Summary: Fast, light, accurate library built for retrieval embedding generation
-Home-page: https://github.com/qdrant/fastembed
 License: Apache License
 Keywords: vector,embedding,neural,search,qdrant,sentence-transformers
 Author: Qdrant Team
@@ -17,20 +16,20 @@ Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3.13
 Requires-Dist: huggingface-hub (>=0.20,<1.0)
 Requires-Dist: loguru (>=0.7.2,<0.8.0)
-Requires-Dist: mmh3 (>=4.1.0,<5.0.0)
+Requires-Dist: mmh3 (>=4.1.0,<6.0.0)
 Requires-Dist: numpy (>=1.21) ; python_version >= "3.10" and python_version < "3.12"
 Requires-Dist: numpy (>=1.21,<2.1.0) ; python_version < "3.10"
-Requires-Dist: numpy (>=1.26) ; python_version >= "3.12" and python_version < "3.13"
+Requires-Dist: numpy (>=1.26) ; python_version == "3.12"
 Requires-Dist: numpy (>=2.1.0) ; python_version >= "3.13"
-Requires-Dist: onnx (>=1.15.0)
 Requires-Dist: onnxruntime-gpu (>1.20.0) ; python_version >= "3.13"
 Requires-Dist: onnxruntime-gpu (>=1.17.0,!=1.20.0) ; python_version >= "3.10" and python_version < "3.13"
 Requires-Dist: onnxruntime-gpu (>=1.17.0,<1.20.0) ; python_version < "3.10"
-Requires-Dist: pillow (>=10.3.0,<11.0.0)
+Requires-Dist: pillow (>=10.3.0,<12.0.0)
 Requires-Dist: py-rust-stemmers (>=0.1.0,<0.2.0)
 Requires-Dist: requests (>=2.31,<3.0)
 Requires-Dist: tokenizers (>=0.15,<1.0)
 Requires-Dist: tqdm (>=4.66,<5.0)
+Project-URL: Homepage, https://github.com/qdrant/fastembed
 Project-URL: Repository, https://github.com/qdrant/fastembed
 Description-Content-Type: text/markdown
 
@@ -99,6 +98,23 @@ embeddings = list(model.embed(documents))
 
 ```
 
+Dense text embedding can also be extended with models which are not in the list of supported models.
+
+```python
+from fastembed import TextEmbedding
+from fastembed.common.model_description import PoolingType, ModelSource
+
+TextEmbedding.add_custom_model(
+    model="intfloat/multilingual-e5-small",
+    pooling=PoolingType.MEAN,
+    normalization=True,
+    sources=ModelSource(hf="intfloat/multilingual-e5-small"),  # can be used with an `url` to load files from a private storage
+    dim=384,
+    model_file="onnx/model.onnx",  # can be used to load an already supported model with another optimization or quantization, e.g. onnx/model_O4.onnx
+)
+model = TextEmbedding(model_name="intfloat/multilingual-e5-small")
+embeddings = list(model.embed(documents))
+```
 
 
 ### 🔱 Sparse text embeddings
@@ -173,6 +189,27 @@ embeddings = list(model.embed(images))
 # ]
 ```
 
+### Late interaction multimodal models (ColPali)
+
+```python
+from fastembed import LateInteractionMultimodalEmbedding
+
+doc_images = [
+    "./path/to/qdrant_pdf_doc_1_screenshot.jpg",
+    "./path/to/colpali_pdf_doc_2_screenshot.jpg",
+]
+
+query = "What is Qdrant?"
+
+model = LateInteractionMultimodalEmbedding(model_name="Qdrant/colpali-v1.3-fp16")
+doc_images_embeddings = list(model.embed_image(doc_images))
+# shape (2, 1030, 128)
+# [array([[-0.03353882, -0.02090454, ..., -0.15576172, -0.07678223]], dtype=float32)]
+query_embedding = model.embed_text(query)
+# shape (1, 20, 128)
+# [array([[-0.00218201,  0.14758301, ...,  -0.02207947,  0.16833496]], dtype=float32)]
+```
+
 ### 🔄 Rerankers
 ```python
 from fastembed.rerank.cross_encoder import TextCrossEncoder

{fastembed_gpu-0.5.0 → fastembed_gpu-0.6.0}/README.md

@@ -63,6 +63,23 @@ embeddings = list(model.embed(documents))
 
 ```
 
+Dense text embedding can also be extended with models which are not in the list of supported models.
+
+```python
+from fastembed import TextEmbedding
+from fastembed.common.model_description import PoolingType, ModelSource
+
+TextEmbedding.add_custom_model(
+    model="intfloat/multilingual-e5-small",
+    pooling=PoolingType.MEAN,
+    normalization=True,
+    sources=ModelSource(hf="intfloat/multilingual-e5-small"),  # can be used with an `url` to load files from a private storage
+    dim=384,
+    model_file="onnx/model.onnx",  # can be used to load an already supported model with another optimization or quantization, e.g. onnx/model_O4.onnx
+)
+model = TextEmbedding(model_name="intfloat/multilingual-e5-small")
+embeddings = list(model.embed(documents))
+```
 
 
 ### 🔱 Sparse text embeddings
@@ -137,6 +154,27 @@ embeddings = list(model.embed(images))
 # ]
 ```
 
+### Late interaction multimodal models (ColPali)
+
+```python
+from fastembed import LateInteractionMultimodalEmbedding
+
+doc_images = [
+    "./path/to/qdrant_pdf_doc_1_screenshot.jpg",
+    "./path/to/colpali_pdf_doc_2_screenshot.jpg",
+]
+
+query = "What is Qdrant?"
+
+model = LateInteractionMultimodalEmbedding(model_name="Qdrant/colpali-v1.3-fp16")
+doc_images_embeddings = list(model.embed_image(doc_images))
+# shape (2, 1030, 128)
+# [array([[-0.03353882, -0.02090454, ..., -0.15576172, -0.07678223]], dtype=float32)]
+query_embedding = model.embed_text(query)
+# shape (1, 20, 128)
+# [array([[-0.00218201,  0.14758301, ...,  -0.02207947,  0.16833496]], dtype=float32)]
+```
+
 ### 🔄 Rerankers
 ```python
 from fastembed.rerank.cross_encoder import TextCrossEncoder

{fastembed_gpu-0.5.0 → fastembed_gpu-0.6.0}/fastembed/__init__.py

@@ -2,6 +2,7 @@ import importlib.metadata
 
 from fastembed.image import ImageEmbedding
 from fastembed.late_interaction import LateInteractionTextEmbedding
+from fastembed.late_interaction_multimodal import LateInteractionMultimodalEmbedding
 from fastembed.sparse import SparseEmbedding, SparseTextEmbedding
 from fastembed.text import TextEmbedding
 
@@ -17,4 +18,5 @@ __all__ = [
     "SparseEmbedding",
     "ImageEmbedding",
     "LateInteractionTextEmbedding",
+    "LateInteractionMultimodalEmbedding",
 ]

fastembed_gpu-0.6.0/fastembed/common/__init__.py

@@ -0,0 +1,3 @@
+from fastembed.common.types import ImageInput, OnnxProvider, PathInput
+
+__all__ = ["OnnxProvider", "ImageInput", "PathInput"]

fastembed_gpu-0.6.0/fastembed/common/model_description.py

@@ -0,0 +1,47 @@
+from dataclasses import dataclass, field
+from enum import Enum
+from typing import Optional, Any
+
+
+@dataclass(frozen=True)
+class ModelSource:
+    hf: Optional[str] = None
+    url: Optional[str] = None
+
+    def __post_init__(self) -> None:
+        if self.hf is None and self.url is None:
+            raise ValueError(
+                f"At least one source should be set, current sources: hf={self.hf}, url={self.url}"
+            )
+
+
+@dataclass(frozen=True)
+class BaseModelDescription:
+    model: str
+    sources: ModelSource
+    model_file: str
+    description: str
+    license: str
+    size_in_GB: float
+    additional_files: list[str] = field(default_factory=list)
+
+
+@dataclass(frozen=True)
+class DenseModelDescription(BaseModelDescription):
+    dim: Optional[int] = None
+    tasks: Optional[dict[str, Any]] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        assert self.dim is not None, "dim is required for dense model description"
+
+
+@dataclass(frozen=True)
+class SparseModelDescription(BaseModelDescription):
+    requires_idf: Optional[bool] = None
+    vocab_size: Optional[int] = None
+
+
+class PoolingType(str, Enum):
+    CLS = "CLS"
+    MEAN = "MEAN"
+    DISABLED = "DISABLED"

{fastembed_gpu-0.5.0 → fastembed_gpu-0.6.0}/fastembed/common/model_management.py

@@ -1,12 +1,14 @@
 import os
 import time
+import json
 import shutil
 import tarfile
 from pathlib import Path
-from typing import Any, Optional
+from typing import Any, Optional, Union, TypeVar, Generic
 
 import requests
-from huggingface_hub import snapshot_download
+from huggingface_hub import snapshot_download, model_info, list_repo_tree
+from huggingface_hub.hf_api import RepoFile
 from huggingface_hub.utils import (
     RepositoryNotFoundError,
     disable_progress_bars,
@@ -14,20 +16,54 @@ from huggingface_hub.utils import (
 )
 from loguru import logger
 from tqdm import tqdm
+from fastembed.common.model_description import BaseModelDescription
 
+T = TypeVar("T", bound=BaseModelDescription)
+
+
+class ModelManagement(Generic[T]):
+    METADATA_FILE = "files_metadata.json"
 
-class ModelManagement:
     @classmethod
     def list_supported_models(cls) -> list[dict[str, Any]]:
         """Lists the supported models.
 
         Returns:
-            list[dict[str, Any]]: A list of dictionaries containing the model information.
+            list[T]: A list of dictionaries containing the model information.
         """
         raise NotImplementedError()
 
     @classmethod
-    def _get_model_description(cls, model_name: str) -> dict[str, Any]:
+    def add_custom_model(
+        cls,
+        *args: Any,
+        **kwargs: Any,
+    ) -> None:
+        """Add a custom model to the existing embedding classes based on the passed model descriptions
+
+        Model description dict should contain the fields same as in one of the model descriptions presented
+         in fastembed.common.model_description
+
+         E.g. for BaseModelDescription:
+              model: str
+              sources: ModelSource
+              model_file: str
+              description: str
+              license: str
+              size_in_GB: float
+              additional_files: list[str]
+
+        Returns:
+            None
+        """
+        raise NotImplementedError()
+
+    @classmethod
+    def _list_supported_models(cls) -> list[T]:
+        raise NotImplementedError()
+
+    @classmethod
+    def _get_model_description(cls, model_name: str) -> T:
         """
         Gets the model description from the model_name.
 
@@ -38,10 +74,10 @@ class ModelManagement:
             ValueError: If the model_name is not supported.
 
         Returns:
-            dict[str, Any]: The model description.
+            T: The model description.
         """
-        for model in cls.list_supported_models():
-            if model_name.lower() == model["model"].lower():
+        for model in cls._list_supported_models():
+            if model_name.lower() == model.model.lower():
                 return model
 
         raise ValueError(f"Model {model_name} is not supported in {cls.__name__}.")
@@ -98,21 +134,74 @@ class ModelManagement:
         cls,
         hf_source_repo: str,
         cache_dir: str,
-        extra_patterns: Optional[list[str]] = None,
+        extra_patterns: list[str],
         local_files_only: bool = False,
-        **kwargs,
+        **kwargs: Any,
     ) -> str:
         """
         Downloads a model from HuggingFace Hub.
         Args:
             hf_source_repo (str): Name of the model on HuggingFace Hub, e.g. "qdrant/all-MiniLM-L6-v2-onnx".
             cache_dir (Optional[str]): The path to the cache directory.
-            extra_patterns (Optional[list[str]]): extra patterns to allow in the snapshot download, typically
+            extra_patterns (list[str]): extra patterns to allow in the snapshot download, typically
                 includes the required model files.
             local_files_only (bool, optional): Whether to only use local files. Defaults to False.
         Returns:
             Path: The path to the model directory.
         """
+
+        def _verify_files_from_metadata(
+            model_dir: Path, stored_metadata: dict[str, Any], repo_files: list[RepoFile]
+        ) -> bool:
+            try:
+                for rel_path, meta in stored_metadata.items():
+                    file_path = model_dir / rel_path
+
+                    if not file_path.exists():
+                        return False
+
+                    if repo_files:  # online verification
+                        file_info = next((f for f in repo_files if f.path == file_path.name), None)
+                        if (
+                            not file_info
+                            or file_info.size != meta["size"]
+                            or file_info.blob_id != meta["blob_id"]
+                        ):
+                            return False
+
+                    else:  # offline verification
+                        if file_path.stat().st_size != meta["size"]:
+                            return False
+                return True
+            except (OSError, KeyError) as e:
+                logger.error(f"Error verifying files: {str(e)}")
+                return False
+
+        def _collect_file_metadata(
+            model_dir: Path, repo_files: list[RepoFile]
+        ) -> dict[str, dict[str, Union[int, str]]]:
+            meta: dict[str, dict[str, Union[int, str]]] = {}
+            file_info_map = {f.path: f for f in repo_files}
+            for file_path in model_dir.rglob("*"):
+                if file_path.is_file() and file_path.name != cls.METADATA_FILE:
+                    repo_file = file_info_map.get(file_path.name)
+                    if repo_file:
+                        meta[str(file_path.relative_to(model_dir))] = {
+                            "size": repo_file.size,
+                            "blob_id": repo_file.blob_id,
+                        }
+            return meta
+
+        def _save_file_metadata(
+            model_dir: Path, meta: dict[str, dict[str, Union[int, str]]]
+        ) -> None:
+            try:
+                if not model_dir.exists():
+                    model_dir.mkdir(parents=True, exist_ok=True)
+                (model_dir / cls.METADATA_FILE).write_text(json.dumps(meta))
+            except (OSError, ValueError) as e:
+                logger.warning(f"Error saving metadata: {str(e)}")
+
         allow_patterns = [
             "config.json",
             "tokenizer.json",
@@ -120,16 +209,59 @@ class ModelManagement:
             "special_tokens_map.json",
             "preprocessor_config.json",
         ]
-        if extra_patterns is not None:
-            allow_patterns.extend(extra_patterns)
+
+        allow_patterns.extend(extra_patterns)
 
         snapshot_dir = Path(cache_dir) / f"models--{hf_source_repo.replace('/', '--')}"
-        is_cached = snapshot_dir.exists()
+        metadata_file = snapshot_dir / cls.METADATA_FILE
 
-        if is_cached:
+        if local_files_only:
+            disable_progress_bars()
+            if metadata_file.exists():
+                metadata = json.loads(metadata_file.read_text())
+                verified = _verify_files_from_metadata(snapshot_dir, metadata, repo_files=[])
+                if not verified:
+                    logger.warning(
+                        "Local file sizes do not match the metadata."
+                    )  # do not raise, still make an attempt to load the model
+            else:
+                logger.warning(
+                    "Metadata file not found. Proceeding without checking local files."
+                )  # if users have downloaded models from hf manually, or they're updating from previous versions of
+                # fastembed
+            result = snapshot_download(
+                repo_id=hf_source_repo,
+                allow_patterns=allow_patterns,
+                cache_dir=cache_dir,
+                local_files_only=local_files_only,
+                **kwargs,
+            )
+            return result
+
+        repo_revision = model_info(hf_source_repo).sha
+        repo_tree = list(list_repo_tree(hf_source_repo, revision=repo_revision, repo_type="model"))
+
+        allowed_extensions = {".json", ".onnx", ".txt"}
+        repo_files = (
+            [
+                f
+                for f in repo_tree
+                if isinstance(f, RepoFile) and Path(f.path).suffix in allowed_extensions
+            ]
+            if repo_tree
+            else []
+        )
+
+        verified_metadata = False
+
+        if snapshot_dir.exists() and metadata_file.exists():
+            metadata = json.loads(metadata_file.read_text())
+            verified_metadata = _verify_files_from_metadata(snapshot_dir, metadata, repo_files)
+
+        if verified_metadata:
             disable_progress_bars()
 
-        return snapshot_download(
+        result = snapshot_download(
             repo_id=hf_source_repo,
             allow_patterns=allow_patterns,
             cache_dir=cache_dir,
@@ -137,8 +269,26 @@ class ModelManagement:
             **kwargs,
         )
 
+        if (
+            not verified_metadata
+        ):  # metadata is not up-to-date, update it and check whether the files have been
+            # downloaded correctly
+            metadata = _collect_file_metadata(snapshot_dir, repo_files)
+
+            download_successful = _verify_files_from_metadata(
+                snapshot_dir, metadata, repo_files=[]
+            )  # offline verification
+            if not download_successful:
+                raise ValueError(
+                    "Files have been corrupted during downloading process. "
+                    "Please check your internet connection and try again."
+                )
+            _save_file_metadata(snapshot_dir, metadata)
+
+        return result
+
     @classmethod
-    def decompress_to_cache(cls, targz_path: str, cache_dir: str):
+    def decompress_to_cache(cls, targz_path: str, cache_dir: str) -> str:
         """
         Decompresses a .tar.gz file to a cache directory.
 
@@ -176,7 +326,11 @@ class ModelManagement:
 
     @classmethod
     def retrieve_model_gcs(
-        cls, model_name: str, source_url: str, cache_dir: str, local_files_only: bool = False
+        cls,
+        model_name: str,
+        source_url: str,
+        cache_dir: str,
+        local_files_only: bool = False,
     ) -> Path:
         fast_model_name = f"fast-{model_name.split('/')[-1]}"
         cache_tmp_dir = Path(cache_dir) / "tmp"
@@ -220,14 +374,12 @@ class ModelManagement:
         return model_dir
 
     @classmethod
-    def download_model(
-        cls, model: dict[str, Any], cache_dir: Path, retries: int = 3, **kwargs
-    ) -> Path:
+    def download_model(cls, model: T, cache_dir: str, retries: int = 3, **kwargs: Any) -> Path:
         """
         Downloads a model from HuggingFace Hub or Google Cloud Storage.
 
         Args:
-            model (dict[str, Any]): The model description.
+            model (T): The model description.
                 Example:
                 ```
                 {
@@ -248,23 +400,26 @@ class ModelManagement:
             Path: The path to the downloaded model directory.
         """
         local_files_only = kwargs.get("local_files_only", False)
+        specific_model_path: Optional[str] = kwargs.pop("specific_model_path", None)
+        if specific_model_path:
+            return Path(specific_model_path)
         retries = 1 if local_files_only else retries
-        hf_source = model.get("sources", {}).get("hf")
-        url_source = model.get("sources", {}).get("url")
+        hf_source = model.sources.hf
+        url_source = model.sources.url
 
         sleep = 3.0
         while retries > 0:
             retries -= 1
 
             if hf_source:
-                extra_patterns = [model["model_file"]]
-                extra_patterns.extend(model.get("additional_files", []))
+                extra_patterns = [model.model_file]
+                extra_patterns.extend(model.additional_files)
 
                 try:
                     return Path(
                         cls.download_files_from_huggingface(
                             hf_source,
-                            cache_dir=str(cache_dir),
+                            cache_dir=cache_dir,
                             extra_patterns=extra_patterns,
                             **kwargs,
                         )
@@ -280,8 +435,8 @@ class ModelManagement:
             if url_source or local_files_only:
                 try:
                     return cls.retrieve_model_gcs(
-                        model["model"],
-                        url_source,
+                        model.model,
+                        str(url_source),
                         str(cache_dir),
                         local_files_only=local_files_only,
                     )
@@ -298,4 +453,4 @@ class ModelManagement:
             time.sleep(sleep)
             sleep *= 3
 
-        raise ValueError(f"Could not load model {model['model']} from any source.")
+        raise ValueError(f"Could not load model {model.model} from any source.")

{fastembed_gpu-0.5.0 → fastembed_gpu-0.6.0}/fastembed/common/onnx_model.py

@@ -6,7 +6,10 @@ from typing import Any, Generic, Iterable, Optional, Sequence, Type, TypeVar
 import numpy as np
 import onnxruntime as ort
 
-from fastembed.common.types import OnnxProvider
+from numpy.typing import NDArray
+from tokenizers import Tokenizer
+
+from fastembed.common.types import OnnxProvider, NumpyArray
 from fastembed.parallel_processor import Worker
 
 # Holds type of the embedding result
@@ -15,26 +18,26 @@ T = TypeVar("T")
 
 @dataclass
 class OnnxOutputContext:
-    model_output: np.ndarray
-    attention_mask: Optional[np.ndarray] = None
-    input_ids: Optional[np.ndarray] = None
+    model_output: NumpyArray
+    attention_mask: Optional[NDArray[np.int64]] = None
+    input_ids: Optional[NDArray[np.int64]] = None
 
 
 class OnnxModel(Generic[T]):
     @classmethod
-    def _get_worker_class(cls) -> Type["EmbeddingWorker"]:
+    def _get_worker_class(cls) -> Type["EmbeddingWorker[T]"]:
         raise NotImplementedError("Subclasses must implement this method")
 
     def _post_process_onnx_output(self, output: OnnxOutputContext) -> Iterable[T]:
         raise NotImplementedError("Subclasses must implement this method")
 
     def __init__(self) -> None:
-        self.model = None
-        self.tokenizer = None
+        self.model: Optional[ort.InferenceSession] = None
+        self.tokenizer: Optional[Tokenizer] = None
 
     def _preprocess_onnx_input(
-        self, onnx_input: dict[str, np.ndarray], **kwargs
-    ) -> dict[str, np.ndarray]:
+        self, onnx_input: dict[str, NumpyArray], **kwargs: Any
+    ) -> dict[str, NumpyArray]:
         """
         Preprocess the onnx input.
         """
@@ -70,7 +73,7 @@ class OnnxModel(Generic[T]):
             onnx_providers = ["CPUExecutionProvider"]
 
         available_providers = ort.get_available_providers()
-        requested_provider_names = []
+        requested_provider_names: list[str] = []
         for provider in onnx_providers:
             # check providers available
             provider_name = provider if isinstance(provider, str) else provider[0]
@@ -91,6 +94,7 @@ class OnnxModel(Generic[T]):
             str(model_path), providers=onnx_providers, sess_options=so
         )
         if "CUDAExecutionProvider" in requested_provider_names:
+            assert self.model is not None
             current_providers = self.model.get_providers()
             if "CUDAExecutionProvider" not in current_providers:
                 warnings.warn(
@@ -103,29 +107,29 @@ class OnnxModel(Generic[T]):
     def load_onnx_model(self) -> None:
         raise NotImplementedError("Subclasses must implement this method")
 
-    def onnx_embed(self, *args, **kwargs) -> OnnxOutputContext:
+    def onnx_embed(self, *args: Any, **kwargs: Any) -> OnnxOutputContext:
         raise NotImplementedError("Subclasses must implement this method")
 
 
-class EmbeddingWorker(Worker):
+class EmbeddingWorker(Worker, Generic[T]):
     def init_embedding(
         self,
         model_name: str,
         cache_dir: str,
-        **kwargs,
-    ) -> OnnxModel:
+        **kwargs: Any,
+    ) -> OnnxModel[T]:
         raise NotImplementedError()
 
     def __init__(
         self,
         model_name: str,
         cache_dir: str,
-        **kwargs,
+        **kwargs: Any,
     ):
         self.model = self.init_embedding(model_name, cache_dir, **kwargs)
 
     @classmethod
-    def start(cls, model_name: str, cache_dir: str, **kwargs: Any) -> "EmbeddingWorker":
+    def start(cls, model_name: str, cache_dir: str, **kwargs: Any) -> "EmbeddingWorker[T]":
         return cls(model_name=model_name, cache_dir=cache_dir, **kwargs)
 
     def process(self, items: Iterable[tuple[int, Any]]) -> Iterable[tuple[int, Any]]: