Перейти к содержанию

Providers API

culicidaelab.providers

Data provider implementations for accessing datasets and models.

This package contains classes that implement the BaseProvider interface to interact with various data sources like Hugging Face, Kaggle, etc. Each provider module offers specific logic for downloading datasets and model weights.

Available Classes
  • HuggingFaceProvider: A provider for interacting with the Hugging Face Hub.
__all__ = ['HuggingFaceProvider'] module-attribute
HuggingFaceProvider

Provider for downloading and managing HuggingFace datasets and models.

This class interfaces with the Hugging Face Hub to fetch dataset metadata, download full datasets or specific splits, and download model weights. It uses the core settings object for path resolution and API key access.

Attributes:

Name Type Description
provider_name str

The name of the provider, "huggingface".
settings Settings

The main Settings object for the library.
dataset_url str

The base URL for fetching Hugging Face dataset metadata.
api_key str | None

The Hugging Face API key, if provided.
Source code in culicidaelab\providers\huggingface_provider.py 
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
class HuggingFaceProvider(BaseProvider):
    """Provider for downloading and managing HuggingFace datasets and models.

    This class interfaces with the Hugging Face Hub to fetch dataset metadata,
    download full datasets or specific splits, and download model weights. It uses
    the core settings object for path resolution and API key access.

    Attributes:
        provider_name (str): The name of the provider, "huggingface".
        settings (Settings): The main Settings object for the library.
        dataset_url (str): The base URL for fetching Hugging Face dataset metadata.
        api_key (str | None): The Hugging Face API key, if provided.
    """

    def __init__(self, settings: Settings, dataset_url: str, **kwargs: Any) -> None:
        """Initializes the HuggingFace provider.

        This constructor is called by the `ProviderService`, which injects the
        global `settings` object and unpacks the specific provider's configuration
        (e.g., `dataset_url`) as keyword arguments.

        Args:
            settings (Settings): The main Settings object for the library.
            dataset_url (str): The base URL for fetching Hugging Face dataset metadata.
            **kwargs (Any): Catches other config parameters (e.g., `api_key`).
        """
        super().__init__()
        self.provider_name = "huggingface"
        self.settings = settings
        self.dataset_url = dataset_url
        self.api_key: str | None = kwargs.get("api_key") or self.settings.get_api_key(
            self.provider_name,
        )

    def download_dataset(
        self,
        dataset_name: str,
        save_dir: Path | None = None,
        config_name: str | None = None,
        split: str | None = None,
        **kwargs: Any,
    ) -> Path:
        """Downloads a dataset from HuggingFace.

        Args:
            dataset_name (str): Name of the dataset to download (e.g., "segmentation").
            save_dir (Path | None, optional): Directory to save the dataset.
                Defaults to None, using the path from settings.
            config_name (str | None, optional): Name of the dataset configuration.
                Defaults to None.
            split (str | None, optional): Dataset split to download (e.g., "train").
                Defaults to None.
            **kwargs (Any): Additional keyword arguments to pass to `load_dataset`.

        Returns:
            Path: The path to the downloaded dataset.

        Raises:
            ValueError: If the configuration is missing the `repository` ID.
            RuntimeError: If the download fails.
        """
        save_path = self.settings.get_dataset_path(dataset_name)
        cache_path = str(self.settings.cache_dir / dataset_name)
        if save_dir:
            save_path = save_dir
        dataset_config = self.settings.get_config(f"datasets.{dataset_name}")

        repo_id = dataset_config.repository
        if not repo_id:
            raise ValueError(
                f"Configuration for dataset '{dataset_name}' is missing the 'repository' (repository ID).",
            )

        try:
            if self.api_key:
                downloaded_object = load_dataset(
                    repo_id,
                    name=config_name,
                    split=split,
                    token=self.api_key,
                    cache_dir=cache_path,
                    **kwargs,
                )
            else:
                downloaded_object = load_dataset(
                    repo_id,
                    name=config_name,
                    split=split,
                    cache_dir=cache_path,
                    **kwargs,
                )

            saveable_dataset = None
            if isinstance(downloaded_object, (IterableDataset, IterableDatasetDict)):
                if isinstance(downloaded_object, IterableDataset):
                    saveable_dataset = Dataset.from_list(list(downloaded_object))
                else:
                    materialized_splits = {s_name: list(s_data) for s_name, s_data in downloaded_object.items()}
                    saveable_dataset = DatasetDict(
                        {s_name: Dataset.from_list(data) for s_name, data in materialized_splits.items()},
                    )
            else:
                saveable_dataset = downloaded_object

            if Path(save_path).exists() and Path(save_path).is_dir():
                shutil.rmtree(save_path)

            save_path.mkdir(parents=True, exist_ok=True)

            saveable_dataset.save_to_disk(str(save_path))

            shutil.rmtree(cache_path, ignore_errors=True)

            return save_path

        except Exception as e:
            if Path(save_path).exists() and Path(save_path).is_dir():
                shutil.rmtree(save_path)
            raise RuntimeError(f"Failed to download dataset {repo_id}: {str(e)}") from e

    def download_model_weights(self, repo_id: str, filename: str, local_dir: Path) -> Path:
        """Downloads and caches a specific model weights file from the HuggingFace Hub.

        Checks if the file exists locally. If not, it downloads it from the specified
        repository and saves it to the given local directory.

        Args:
            repo_id (str): The Hugging Face repository ID (e.g., 'user/repo-name').
            filename (str): The specific weights file to download (e.g., 'model.onnx').
            local_dir (Path): The local directory to save the file in.

        Returns:
            Path: The absolute path to the downloaded model weights file.

        Raises:
            ValueError: If `repo_id` or `filename` are not provided.
            RuntimeError: If the download fails for any reason.
        """
        if not repo_id or not filename:
            raise ValueError("'repo_id' and 'filename' must be provided.")

        local_path = (local_dir / filename).resolve()
        cache_dir = self.settings.cache_dir / f"{repo_id.replace('/', '_')}_{filename}"

        if local_path.exists():
            print(f"Weights file found at: {local_path}")
            return local_path

        print(f"Model weights '{filename}' not found locally. Attempting to download from {repo_id}...")

        try:
            local_dir.mkdir(parents=True, exist_ok=True)

            downloaded_path_str = hf_hub_download(
                repo_id=repo_id,
                filename=filename,
                cache_dir=str(cache_dir),
                local_dir=str(local_dir),
                # local_dir_use_symlinks=False,  # Use direct file placement
            )

            downloaded_path = Path(downloaded_path_str)

            # hf_hub_download might place it in a subfolder structure, so we ensure it's moved to the final destination
            if downloaded_path.resolve() != local_path.resolve():
                shutil.move(str(downloaded_path), str(local_path))
                print(f"Moved weights to final destination: {local_path}")
            else:
                print(f"Downloaded weights directly to: {local_path}")

            # Clean up the cache directory if it was used
            if cache_dir.exists():
                shutil.rmtree(cache_dir, ignore_errors=True)

            return local_path

        except Exception as e:
            # Clean up any partial download
            if local_path.exists():
                local_path.unlink()
            raise RuntimeError(
                f"Failed to download weights file '{filename}' from repo '{repo_id}'. Error: {e}",
            ) from e

    def get_dataset_metadata(self, dataset_name: str) -> dict[str, Any]:
        """Gets metadata for a specific dataset from HuggingFace.

        Args:
            dataset_name (str): The name of the dataset to get metadata for.

        Returns:
            dict[str, Any]: The dataset metadata as a dictionary.

        Raises:
            requests.RequestException: If the HTTP request fails.
        """
        url = self.dataset_url.format(dataset_name=dataset_name)
        headers = {"Authorization": f"Bearer {self.api_key}"} if self.api_key else {}

        try:
            response = requests.get(url, headers=headers, timeout=10.0)
            response.raise_for_status()
            return cast(dict[str, Any], response.json())
        except requests.RequestException as e:
            raise requests.RequestException(
                f"Failed to fetch dataset metadata for {dataset_name}: {str(e)}",
            ) from e

    def get_provider_name(self) -> str:
        """Returns the provider's name.

        Returns:
            str: The name of the provider, "huggingface".
        """
        return self.provider_name

    def load_dataset(self, dataset_path: str | Path, **kwargs: Any) -> Any:
        """Loads a dataset from disk.

        This method attempts to load a dataset from the specified path. If a `split`
        name is provided and a corresponding subdirectory exists, it will load
        the split from that subdirectory. Otherwise, it loads the entire dataset
        from the base path.

        Args:
            dataset_path (str | Path): The local path to the dataset,
                typically returned by `download_dataset`.
            **kwargs: Additional keyword arguments to pass to the
                `datasets.load_from_disk` function.

        Returns:
            Any: The loaded dataset, typically a `datasets.Dataset` or
                `datasets.DatasetDict` object.
        """
        return load_from_disk(str(dataset_path), **kwargs)
provider_name = 'huggingface' instance-attribute
settings = settings instance-attribute
dataset_url = dataset_url instance-attribute
api_key: str | None = kwargs.get('api_key') or self.settings.get_api_key(self.provider_name) instance-attribute
__init__(settings: Settings, dataset_url: str, **kwargs: Any) -> None

Initializes the HuggingFace provider.

This constructor is called by the ProviderService, which injects the global settings object and unpacks the specific provider's configuration (e.g., dataset_url) as keyword arguments.

Parameters:

Name Type Description Default
settings Settings

The main Settings object for the library.

 required
dataset_url str

The base URL for fetching Hugging Face dataset metadata.

 required
**kwargs Any

Catches other config parameters (e.g., api_key).

 {}
Source code in culicidaelab\providers\huggingface_provider.py 
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
def __init__(self, settings: Settings, dataset_url: str, **kwargs: Any) -> None:
    """Initializes the HuggingFace provider.

    This constructor is called by the `ProviderService`, which injects the
    global `settings` object and unpacks the specific provider's configuration
    (e.g., `dataset_url`) as keyword arguments.

    Args:
        settings (Settings): The main Settings object for the library.
        dataset_url (str): The base URL for fetching Hugging Face dataset metadata.
        **kwargs (Any): Catches other config parameters (e.g., `api_key`).
    """
    super().__init__()
    self.provider_name = "huggingface"
    self.settings = settings
    self.dataset_url = dataset_url
    self.api_key: str | None = kwargs.get("api_key") or self.settings.get_api_key(
        self.provider_name,
    )
download_dataset(dataset_name: str, save_dir: Path | None = None, config_name: str | None = None, split: str | None = None, **kwargs: Any) -> Path

Downloads a dataset from HuggingFace.

Parameters:

Name Type Description Default
dataset_name str

Name of the dataset to download (e.g., "segmentation").

 required
save_dir Path | None

Directory to save the dataset. Defaults to None, using the path from settings.

 None
config_name str | None

Name of the dataset configuration. Defaults to None.

 None
split str | None

Dataset split to download (e.g., "train"). Defaults to None.

 None
**kwargs Any

Additional keyword arguments to pass to load_dataset.

 {}

Returns:

Name Type Description
Path Path

The path to the downloaded dataset.

Raises:

Type Description
ValueError

If the configuration is missing the repository ID.
RuntimeError

If the download fails.
Source code in culicidaelab\providers\huggingface_provider.py 
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
def download_dataset(
    self,
    dataset_name: str,
    save_dir: Path | None = None,
    config_name: str | None = None,
    split: str | None = None,
    **kwargs: Any,
) -> Path:
    """Downloads a dataset from HuggingFace.

    Args:
        dataset_name (str): Name of the dataset to download (e.g., "segmentation").
        save_dir (Path | None, optional): Directory to save the dataset.
            Defaults to None, using the path from settings.
        config_name (str | None, optional): Name of the dataset configuration.
            Defaults to None.
        split (str | None, optional): Dataset split to download (e.g., "train").
            Defaults to None.
        **kwargs (Any): Additional keyword arguments to pass to `load_dataset`.

    Returns:
        Path: The path to the downloaded dataset.

    Raises:
        ValueError: If the configuration is missing the `repository` ID.
        RuntimeError: If the download fails.
    """
    save_path = self.settings.get_dataset_path(dataset_name)
    cache_path = str(self.settings.cache_dir / dataset_name)
    if save_dir:
        save_path = save_dir
    dataset_config = self.settings.get_config(f"datasets.{dataset_name}")

    repo_id = dataset_config.repository
    if not repo_id:
        raise ValueError(
            f"Configuration for dataset '{dataset_name}' is missing the 'repository' (repository ID).",
        )

    try:
        if self.api_key:
            downloaded_object = load_dataset(
                repo_id,
                name=config_name,
                split=split,
                token=self.api_key,
                cache_dir=cache_path,
                **kwargs,
            )
        else:
            downloaded_object = load_dataset(
                repo_id,
                name=config_name,
                split=split,
                cache_dir=cache_path,
                **kwargs,
            )

        saveable_dataset = None
        if isinstance(downloaded_object, (IterableDataset, IterableDatasetDict)):
            if isinstance(downloaded_object, IterableDataset):
                saveable_dataset = Dataset.from_list(list(downloaded_object))
            else:
                materialized_splits = {s_name: list(s_data) for s_name, s_data in downloaded_object.items()}
                saveable_dataset = DatasetDict(
                    {s_name: Dataset.from_list(data) for s_name, data in materialized_splits.items()},
                )
        else:
            saveable_dataset = downloaded_object

        if Path(save_path).exists() and Path(save_path).is_dir():
            shutil.rmtree(save_path)

        save_path.mkdir(parents=True, exist_ok=True)

        saveable_dataset.save_to_disk(str(save_path))

        shutil.rmtree(cache_path, ignore_errors=True)

        return save_path

    except Exception as e:
        if Path(save_path).exists() and Path(save_path).is_dir():
            shutil.rmtree(save_path)
        raise RuntimeError(f"Failed to download dataset {repo_id}: {str(e)}") from e
download_model_weights(repo_id: str, filename: str, local_dir: Path) -> Path

Downloads and caches a specific model weights file from the HuggingFace Hub.

Checks if the file exists locally. If not, it downloads it from the specified repository and saves it to the given local directory.

Parameters:

Name Type Description Default
repo_id str

The Hugging Face repository ID (e.g., 'user/repo-name').

 required
filename str

The specific weights file to download (e.g., 'model.onnx').

 required
local_dir Path

The local directory to save the file in.

 required

Returns:

Name Type Description
Path Path

The absolute path to the downloaded model weights file.

Raises:

Type Description
ValueError

If repo_id or filename are not provided.
RuntimeError

If the download fails for any reason.
Source code in culicidaelab\providers\huggingface_provider.py 
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
def download_model_weights(self, repo_id: str, filename: str, local_dir: Path) -> Path:
    """Downloads and caches a specific model weights file from the HuggingFace Hub.

    Checks if the file exists locally. If not, it downloads it from the specified
    repository and saves it to the given local directory.

    Args:
        repo_id (str): The Hugging Face repository ID (e.g., 'user/repo-name').
        filename (str): The specific weights file to download (e.g., 'model.onnx').
        local_dir (Path): The local directory to save the file in.

    Returns:
        Path: The absolute path to the downloaded model weights file.

    Raises:
        ValueError: If `repo_id` or `filename` are not provided.
        RuntimeError: If the download fails for any reason.
    """
    if not repo_id or not filename:
        raise ValueError("'repo_id' and 'filename' must be provided.")

    local_path = (local_dir / filename).resolve()
    cache_dir = self.settings.cache_dir / f"{repo_id.replace('/', '_')}_{filename}"

    if local_path.exists():
        print(f"Weights file found at: {local_path}")
        return local_path

    print(f"Model weights '{filename}' not found locally. Attempting to download from {repo_id}...")

    try:
        local_dir.mkdir(parents=True, exist_ok=True)

        downloaded_path_str = hf_hub_download(
            repo_id=repo_id,
            filename=filename,
            cache_dir=str(cache_dir),
            local_dir=str(local_dir),
            # local_dir_use_symlinks=False,  # Use direct file placement
        )

        downloaded_path = Path(downloaded_path_str)

        # hf_hub_download might place it in a subfolder structure, so we ensure it's moved to the final destination
        if downloaded_path.resolve() != local_path.resolve():
            shutil.move(str(downloaded_path), str(local_path))
            print(f"Moved weights to final destination: {local_path}")
        else:
            print(f"Downloaded weights directly to: {local_path}")

        # Clean up the cache directory if it was used
        if cache_dir.exists():
            shutil.rmtree(cache_dir, ignore_errors=True)

        return local_path

    except Exception as e:
        # Clean up any partial download
        if local_path.exists():
            local_path.unlink()
        raise RuntimeError(
            f"Failed to download weights file '{filename}' from repo '{repo_id}'. Error: {e}",
        ) from e
get_dataset_metadata(dataset_name: str) -> dict[str, Any]

Gets metadata for a specific dataset from HuggingFace.

Parameters:

Name Type Description Default
dataset_name str

The name of the dataset to get metadata for.

 required

Returns:

Type Description
dict[str, Any]

dict[str, Any]: The dataset metadata as a dictionary.

Raises:

Type Description
RequestException

If the HTTP request fails.
Source code in culicidaelab\providers\huggingface_provider.py 
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
def get_dataset_metadata(self, dataset_name: str) -> dict[str, Any]:
    """Gets metadata for a specific dataset from HuggingFace.

    Args:
        dataset_name (str): The name of the dataset to get metadata for.

    Returns:
        dict[str, Any]: The dataset metadata as a dictionary.

    Raises:
        requests.RequestException: If the HTTP request fails.
    """
    url = self.dataset_url.format(dataset_name=dataset_name)
    headers = {"Authorization": f"Bearer {self.api_key}"} if self.api_key else {}

    try:
        response = requests.get(url, headers=headers, timeout=10.0)
        response.raise_for_status()
        return cast(dict[str, Any], response.json())
    except requests.RequestException as e:
        raise requests.RequestException(
            f"Failed to fetch dataset metadata for {dataset_name}: {str(e)}",
        ) from e
get_provider_name() -> str

Returns the provider's name.

Returns:

Name Type Description
str str

The name of the provider, "huggingface".
Source code in culicidaelab\providers\huggingface_provider.py 
241
242
243
244
245
246
247
def get_provider_name(self) -> str:
    """Returns the provider's name.

    Returns:
        str: The name of the provider, "huggingface".
    """
    return self.provider_name
load_dataset(dataset_path: str | Path, **kwargs: Any) -> Any

Loads a dataset from disk.

This method attempts to load a dataset from the specified path. If a split name is provided and a corresponding subdirectory exists, it will load the split from that subdirectory. Otherwise, it loads the entire dataset from the base path.

Parameters:

Name Type Description Default
dataset_path str | Path

The local path to the dataset, typically returned by download_dataset.

 required
**kwargs Any

Additional keyword arguments to pass to the datasets.load_from_disk function.

 {}

Returns:

Name Type Description
Any Any

The loaded dataset, typically a datasets.Dataset or datasets.DatasetDict object.
Source code in culicidaelab\providers\huggingface_provider.py 
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
def load_dataset(self, dataset_path: str | Path, **kwargs: Any) -> Any:
    """Loads a dataset from disk.

    This method attempts to load a dataset from the specified path. If a `split`
    name is provided and a corresponding subdirectory exists, it will load
    the split from that subdirectory. Otherwise, it loads the entire dataset
    from the base path.

    Args:
        dataset_path (str | Path): The local path to the dataset,
            typically returned by `download_dataset`.
        **kwargs: Additional keyword arguments to pass to the
            `datasets.load_from_disk` function.

    Returns:
        Any: The loaded dataset, typically a `datasets.Dataset` or
            `datasets.DatasetDict` object.
    """
    return load_from_disk(str(dataset_path), **kwargs)

selection:

members: true