huggingface-hub 0.34.5__py3-none-any.whl → 0.35.0__py3-none-any.whl

huggingface_hub/hf_file_system.py

@@ -896,7 +896,7 @@ class HfFileSystem(fsspec.AbstractFileSystem):
                     repo_type=resolve_remote_path.repo_type,
                     endpoint=self.endpoint,
                 ),
-                temp_file=outfile,
+                temp_file=outfile,  # type: ignore[arg-type]
                 displayed_filename=rpath,
                 expected_size=expected_size,
                 resume_size=0,
@@ -1069,6 +1069,7 @@ class HfFileSystemStreamFile(fsspec.spec.AbstractBufferedFile):
             )
             hf_raise_for_status(self.response)
         try:
+            self.response.raw.decode_content = True
             out = self.response.raw.read(*read_args)
         except Exception:
             self.response.close()
@@ -1091,6 +1092,7 @@ class HfFileSystemStreamFile(fsspec.spec.AbstractBufferedFile):
             )
             hf_raise_for_status(self.response)
             try:
+                self.response.raw.decode_content = True
                 out = self.response.raw.read(*read_args)
             except Exception:
                 self.response.close()

huggingface_hub/hub_mixin.py

@@ -266,12 +266,14 @@ class ModelHubMixin:
         if pipeline_tag is not None:
             info.model_card_data.pipeline_tag = pipeline_tag
         if tags is not None:
+            normalized_tags = list(tags)
             if info.model_card_data.tags is not None:
-                info.model_card_data.tags.extend(tags)
+                info.model_card_data.tags.extend(normalized_tags)
             else:
-                info.model_card_data.tags = tags
+                info.model_card_data.tags = normalized_tags
 
-        info.model_card_data.tags = sorted(set(info.model_card_data.tags))
+        if info.model_card_data.tags is not None:
+            info.model_card_data.tags = sorted(set(info.model_card_data.tags))
 
         # Handle encoders/decoders for args
         cls._hub_mixin_coders = coders or {}

huggingface_hub/inference/_client.py

@@ -45,7 +45,6 @@ from huggingface_hub.errors import BadRequestError, InferenceTimeoutError
 from huggingface_hub.inference._common import (
     TASKS_EXPECTING_IMAGES,
     ContentT,
-    ModelStatus,
     RequestParameters,
     _b64_encode,
     _b64_to_image,
@@ -54,7 +53,6 @@ from huggingface_hub.inference._common import (
     _bytes_to_list,
     _get_unsupported_text_generation_kwargs,
     _import_numpy,
-    _open_as_binary,
     _set_unsupported_text_generation_kwargs,
     _stream_chat_completion_response,
     _stream_text_generation_response,
@@ -105,7 +103,6 @@ from huggingface_hub.inference._generated.types import (
 from huggingface_hub.inference._providers import PROVIDER_OR_POLICY_T, get_provider_helper
 from huggingface_hub.utils import build_hf_headers, get_session, hf_raise_for_status
 from huggingface_hub.utils._auth import get_token
-from huggingface_hub.utils._deprecation import _deprecate_method
 
 
 if TYPE_CHECKING:
@@ -133,7 +130,7 @@ class InferenceClient:
             Note: for better compatibility with OpenAI's client, `model` has been aliased as `base_url`. Those 2
             arguments are mutually exclusive. If a URL is passed as `model` or `base_url` for chat completion, the `(/v1)/chat/completions` suffix path will be appended to the URL.
         provider (`str`, *optional*):
-            Name of the provider to use for inference. Can be `"black-forest-labs"`, `"cerebras"`, `"cohere"`, `"fal-ai"`, `"featherless-ai"`, `"fireworks-ai"`, `"groq"`, `"hf-inference"`, `"hyperbolic"`, `"nebius"`, `"novita"`, `"nscale"`, `"openai"`, `"replicate"`, `"sambanova"`, `"scaleway"` or `"together"`.
+            Name of the provider to use for inference. Can be `"black-forest-labs"`, `"cerebras"`, `"cohere"`, `"fal-ai"`, `"featherless-ai"`, `"fireworks-ai"`, `"groq"`, `"hf-inference"`, `"hyperbolic"`, `"nebius"`, `"novita"`, `"nscale"`, `"openai"`, `publicai`, `"replicate"`, `"sambanova"`, `"scaleway"` or `"together"`.
             Defaults to "auto" i.e. the first of the providers available for the model, sorted by the user's order in https://hf.co/settings/inference-providers.
             If model is a URL or `base_url` is passed, then `provider` is not used.
         token (`str`, *optional*):
@@ -259,21 +256,20 @@ class InferenceClient:
         if request_parameters.task in TASKS_EXPECTING_IMAGES and "Accept" not in request_parameters.headers:
             request_parameters.headers["Accept"] = "image/png"
 
-        with _open_as_binary(request_parameters.data) as data_as_binary:
-            try:
-                response = get_session().post(
-                    request_parameters.url,
-                    json=request_parameters.json,
-                    data=data_as_binary,
-                    headers=request_parameters.headers,
-                    cookies=self.cookies,
-                    timeout=self.timeout,
-                    stream=stream,
-                    proxies=self.proxies,
-                )
-            except TimeoutError as error:
-                # Convert any `TimeoutError` to a `InferenceTimeoutError`
-                raise InferenceTimeoutError(f"Inference call timed out: {request_parameters.url}") from error  # type: ignore
+        try:
+            response = get_session().post(
+                request_parameters.url,
+                json=request_parameters.json,
+                data=request_parameters.data,
+                headers=request_parameters.headers,
+                cookies=self.cookies,
+                timeout=self.timeout,
+                stream=stream,
+                proxies=self.proxies,
+            )
+        except TimeoutError as error:
+            # Convert any `TimeoutError` to a `InferenceTimeoutError`
+            raise InferenceTimeoutError(f"Inference call timed out: {request_parameters.url}") from error  # type: ignore
 
         try:
             hf_raise_for_status(response)
@@ -1462,8 +1458,8 @@ class InferenceClient:
             api_key=self.token,
         )
         response = self._inner_post(request_parameters)
-        output = ImageToTextOutput.parse_obj(response)
-        return output[0] if isinstance(output, list) else output
+        output_list: List[ImageToTextOutput] = ImageToTextOutput.parse_obj_as_list(response)
+        return output_list[0]
 
     def object_detection(
         self, image: ContentT, *, model: Optional[str] = None, threshold: Optional[float] = None
@@ -3273,101 +3269,6 @@ class InferenceClient:
         response = self._inner_post(request_parameters)
         return ZeroShotImageClassificationOutputElement.parse_obj_as_list(response)
 
-    @_deprecate_method(
-        version="0.35.0",
-        message=(
-            "HF Inference API is getting revamped and will only support warm models in the future (no cold start allowed)."
-            " Use `HfApi.list_models(..., inference_provider='...')` to list warm models per provider."
-        ),
-    )
-    def list_deployed_models(
-        self, frameworks: Union[None, str, Literal["all"], List[str]] = None
-    ) -> Dict[str, List[str]]:
-        """
-        List models deployed on the HF Serverless Inference API service.
-
-        This helper checks deployed models framework by framework. By default, it will check the 4 main frameworks that
-        are supported and account for 95% of the hosted models. However, if you want a complete list of models you can
-        specify `frameworks="all"` as input. Alternatively, if you know before-hand which framework you are interested
-        in, you can also restrict to search to this one (e.g. `frameworks="text-generation-inference"`). The more
-        frameworks are checked, the more time it will take.
-
-        <Tip warning={true}>
-
-        This endpoint method does not return a live list of all models available for the HF Inference API service.
-        It searches over a cached list of models that were recently available and the list may not be up to date.
-        If you want to know the live status of a specific model, use [`~InferenceClient.get_model_status`].
-
-        </Tip>
-
-        <Tip>
-
-        This endpoint method is mostly useful for discoverability. If you already know which model you want to use and want to
-        check its availability, you can directly use [`~InferenceClient.get_model_status`].
-
-        </Tip>
-
-        Args:
-            frameworks (`Literal["all"]` or `List[str]` or `str`, *optional*):
-                The frameworks to filter on. By default only a subset of the available frameworks are tested. If set to
-                "all", all available frameworks will be tested. It is also possible to provide a single framework or a
-                custom set of frameworks to check.
-
-        Returns:
-            `Dict[str, List[str]]`: A dictionary mapping task names to a sorted list of model IDs.
-
-        Example:
-        ```python
-        >>> from huggingface_hub import InferenceClient
-        >>> client = InferenceClient()
-
-        # Discover zero-shot-classification models currently deployed
-        >>> models = client.list_deployed_models()
-        >>> models["zero-shot-classification"]
-        ['Narsil/deberta-large-mnli-zero-cls', 'facebook/bart-large-mnli', ...]
-
-        # List from only 1 framework
-        >>> client.list_deployed_models("text-generation-inference")
-        {'text-generation': ['bigcode/starcoder', 'meta-llama/Llama-2-70b-chat-hf', ...], ...}
-        ```
-        """
-        if self.provider != "hf-inference":
-            raise ValueError(f"Listing deployed models is not supported on '{self.provider}'.")
-
-        # Resolve which frameworks to check
-        if frameworks is None:
-            frameworks = constants.MAIN_INFERENCE_API_FRAMEWORKS
-        elif frameworks == "all":
-            frameworks = constants.ALL_INFERENCE_API_FRAMEWORKS
-        elif isinstance(frameworks, str):
-            frameworks = [frameworks]
-        frameworks = list(set(frameworks))
-
-        # Fetch them iteratively
-        models_by_task: Dict[str, List[str]] = {}
-
-        def _unpack_response(framework: str, items: List[Dict]) -> None:
-            for model in items:
-                if framework == "sentence-transformers":
-                    # Model running with the `sentence-transformers` framework can work with both tasks even if not
-                    # branded as such in the API response
-                    models_by_task.setdefault("feature-extraction", []).append(model["model_id"])
-                    models_by_task.setdefault("sentence-similarity", []).append(model["model_id"])
-                else:
-                    models_by_task.setdefault(model["task"], []).append(model["model_id"])
-
-        for framework in frameworks:
-            response = get_session().get(
-                f"{constants.INFERENCE_ENDPOINT}/framework/{framework}", headers=build_hf_headers(token=self.token)
-            )
-            hf_raise_for_status(response)
-            _unpack_response(framework, response.json())
-
-        # Sort alphabetically for discoverability and return
-        for task, models in models_by_task.items():
-            models_by_task[task] = sorted(set(models), key=lambda x: x.lower())
-        return models_by_task
-
     def get_endpoint_info(self, *, model: Optional[str] = None) -> Dict[str, Any]:
         """
         Get information about the deployed endpoint.
@@ -3431,7 +3332,6 @@ class InferenceClient:
         Check the health of the deployed endpoint.
 
         Health check is only available with Inference Endpoints powered by Text-Generation-Inference (TGI) or Text-Embedding-Inference (TEI).
-        For Inference API, please use [`InferenceClient.get_model_status`] instead.
 
         Args:
             model (`str`, *optional*):
@@ -3455,75 +3355,12 @@ class InferenceClient:
         if model is None:
             raise ValueError("Model id not provided.")
         if not model.startswith(("http://", "https://")):
-            raise ValueError(
-                "Model must be an Inference Endpoint URL. For serverless Inference API, please use `InferenceClient.get_model_status`."
-            )
+            raise ValueError("Model must be an Inference Endpoint URL.")
         url = model.rstrip("/") + "/health"
 
         response = get_session().get(url, headers=build_hf_headers(token=self.token))
         return response.status_code == 200
 
-    @_deprecate_method(
-        version="0.35.0",
-        message=(
-            "HF Inference API is getting revamped and will only support warm models in the future (no cold start allowed)."
-            " Use `HfApi.model_info` to get the model status both with HF Inference API and external providers."
-        ),
-    )
-    def get_model_status(self, model: Optional[str] = None) -> ModelStatus:
-        """
-        Get the status of a model hosted on the HF Inference API.
-
-        <Tip>
-
-        This endpoint is mostly useful when you already know which model you want to use and want to check its
-        availability. If you want to discover already deployed models, you should rather use [`~InferenceClient.list_deployed_models`].
-
-        </Tip>
-
-        Args:
-            model (`str`, *optional*):
-                Identifier of the model for witch the status gonna be checked. If model is not provided,
-                the model associated with this instance of [`InferenceClient`] will be used. Only HF Inference API service can be checked so the
-                identifier cannot be a URL.
-
-
-        Returns:
-            [`ModelStatus`]: An instance of ModelStatus dataclass, containing information,
-                         about the state of the model: load, state, compute type and framework.
-
-        Example:
-        ```py
-        >>> from huggingface_hub import InferenceClient
-        >>> client = InferenceClient()
-        >>> client.get_model_status("meta-llama/Meta-Llama-3-8B-Instruct")
-        ModelStatus(loaded=True, state='Loaded', compute_type='gpu', framework='text-generation-inference')
-        ```
-        """
-        if self.provider != "hf-inference":
-            raise ValueError(f"Getting model status is not supported on '{self.provider}'.")
-
-        model = model or self.model
-        if model is None:
-            raise ValueError("Model id not provided.")
-        if model.startswith("https://"):
-            raise NotImplementedError("Model status is only available for Inference API endpoints.")
-        url = f"{constants.INFERENCE_ENDPOINT}/status/{model}"
-
-        response = get_session().get(url, headers=build_hf_headers(token=self.token))
-        hf_raise_for_status(response)
-        response_data = response.json()
-
-        if "error" in response_data:
-            raise ValueError(response_data["error"])
-
-        return ModelStatus(
-            loaded=response_data["loaded"],
-            state=response_data["state"],
-            compute_type=response_data["compute_type"],
-            framework=response_data["framework"],
-        )
-
     @property
     def chat(self) -> "ProxyClientChat":
         return ProxyClientChat(self)

huggingface_hub/inference/_common.py

@@ -19,7 +19,6 @@ import io
 import json
 import logging
 import mimetypes
-from contextlib import contextmanager
 from dataclasses import dataclass
 from pathlib import Path
 from typing import (
@@ -27,9 +26,7 @@ from typing import (
     Any,
     AsyncIterable,
     BinaryIO,
-    ContextManager,
     Dict,
-    Generator,
     Iterable,
     List,
     Literal,
@@ -61,8 +58,7 @@ if TYPE_CHECKING:
 # TYPES
 UrlT = str
 PathT = Union[str, Path]
-BinaryT = Union[bytes, BinaryIO]
-ContentT = Union[BinaryT, PathT, UrlT, "Image"]
+ContentT = Union[bytes, BinaryIO, PathT, UrlT, "Image", bytearray, memoryview]
 
 # Use to set a Accept: image/png header
 TASKS_EXPECTING_IMAGES = {"text-to-image", "image-to-image"}
@@ -76,39 +72,33 @@ class RequestParameters:
     task: str
     model: Optional[str]
     json: Optional[Union[str, Dict, List]]
-    data: Optional[ContentT]
+    data: Optional[bytes]
     headers: Dict[str, Any]
 
 
-# Add dataclass for ModelStatus. We use this dataclass in get_model_status function.
-@dataclass
-class ModelStatus:
+class MimeBytes(bytes):
     """
-    This Dataclass represents the model status in the HF Inference API.
-
-    Args:
-        loaded (`bool`):
-            If the model is currently loaded into HF's Inference API. Models
-            are loaded on-demand, leading to the user's first request taking longer.
-            If a model is loaded, you can be assured that it is in a healthy state.
-        state (`str`):
-            The current state of the model. This can be 'Loaded', 'Loadable', 'TooBig'.
-            If a model's state is 'Loadable', it's not too big and has a supported
-            backend. Loadable models are automatically loaded when the user first
-            requests inference on the endpoint. This means it is transparent for the
-            user to load a model, except that the first call takes longer to complete.
-        compute_type (`Dict`):
-            Information about the compute resource the model is using or will use, such as 'gpu' type and number of
-            replicas.
-        framework (`str`):
-            The name of the framework that the model was built with, such as 'transformers'
-            or 'text-generation-inference'.
+    A bytes object with a mime type.
+    To be returned by `_prepare_payload_open_as_mime_bytes` in subclasses.
+
+    Example:
+    ```python
+        >>> b = MimeBytes(b"hello", "text/plain")
+        >>> isinstance(b, bytes)
+        True
+        >>> b.mime_type
+        'text/plain'
+    ```
     """
 
-    loaded: bool
-    state: str
-    compute_type: Dict
-    framework: str
+    mime_type: Optional[str]
+
+    def __new__(cls, data: bytes, mime_type: Optional[str] = None):
+        obj = super().__new__(cls, data)
+        obj.mime_type = mime_type
+        if isinstance(data, MimeBytes) and mime_type is None:
+            obj.mime_type = data.mime_type
+        return obj
 
 
 ## IMPORT UTILS
@@ -148,31 +138,49 @@ def _import_pil_image():
 
 
 @overload
-def _open_as_binary(
-    content: ContentT,
-) -> ContextManager[BinaryT]: ...  # means "if input is not None, output is not None"
+def _open_as_mime_bytes(content: ContentT) -> MimeBytes: ...  # means "if input is not None, output is not None"
 
 
 @overload
-def _open_as_binary(
-    content: Literal[None],
-) -> ContextManager[Literal[None]]: ...  # means "if input is None, output is None"
+def _open_as_mime_bytes(content: Literal[None]) -> Literal[None]: ...  # means "if input is None, output is None"
 
 
-@contextmanager  # type: ignore
-def _open_as_binary(content: Optional[ContentT]) -> Generator[Optional[BinaryT], None, None]:
+def _open_as_mime_bytes(content: Optional[ContentT]) -> Optional[MimeBytes]:
     """Open `content` as a binary file, either from a URL, a local path, raw bytes, or a PIL Image.
 
     Do nothing if `content` is None.
-
-    TODO: handle base64 as input
     """
+    # If content is None, yield None
+    if content is None:
+        return None
+
+    # If content is bytes, return it
+    if isinstance(content, bytes):
+        return MimeBytes(content)
+
+    # If content is raw binary data (bytearray, memoryview)
+    if isinstance(content, (bytearray, memoryview)):
+        return MimeBytes(bytes(content))
+
+    # If content is a binary file-like object
+    if hasattr(content, "read"):  # duck-typing instead of isinstance(content, BinaryIO)
+        logger.debug("Reading content from BinaryIO")
+        data = content.read()
+        mime_type = mimetypes.guess_type(content.name)[0] if hasattr(content, "name") else None
+        if isinstance(data, str):
+            raise TypeError("Expected binary stream (bytes), but got text stream")
+        return MimeBytes(data, mime_type=mime_type)
+
     # If content is a string => must be either a URL or a path
     if isinstance(content, str):
         if content.startswith("https://") or content.startswith("http://"):
             logger.debug(f"Downloading content from {content}")
-            yield get_session().get(content).content  # TODO: retrieve as stream and pipe to post request ?
-            return
+            response = get_session().get(content)
+            mime_type = response.headers.get("Content-Type")
+            if mime_type is None:
+                mime_type = mimetypes.guess_type(content)[0]
+            return MimeBytes(response.content, mime_type=mime_type)
+
         content = Path(content)
         if not content.exists():
             raise FileNotFoundError(
@@ -183,9 +191,7 @@ def _open_as_binary(content: Optional[ContentT]) -> Generator[Optional[BinaryT],
     # If content is a Path => open it
     if isinstance(content, Path):
         logger.debug(f"Opening content from {content}")
-        with content.open("rb") as f:
-            yield f
-        return
+        return MimeBytes(content.read_bytes(), mime_type=mimetypes.guess_type(content)[0])
 
     # If content is a PIL Image => convert to bytes
     if is_pillow_available():
@@ -194,38 +200,37 @@ def _open_as_binary(content: Optional[ContentT]) -> Generator[Optional[BinaryT],
         if isinstance(content, Image.Image):
             logger.debug("Converting PIL Image to bytes")
             buffer = io.BytesIO()
-            content.save(buffer, format=content.format or "PNG")
-            yield buffer.getvalue()
-            return
+            format = content.format or "PNG"
+            content.save(buffer, format=format)
+            return MimeBytes(buffer.getvalue(), mime_type=f"image/{format.lower()}")
 
-    # Otherwise: already a file-like object or None
-    yield content  # type: ignore
+    # If nothing matched, raise error
+    raise TypeError(
+        f"Unsupported content type: {type(content)}. "
+        "Expected one of: bytes, bytearray, BinaryIO, memoryview, Path, str (URL or file path), or PIL.Image.Image."
+    )
 
 
 def _b64_encode(content: ContentT) -> str:
     """Encode a raw file (image, audio) into base64. Can be bytes, an opened file, a path or a URL."""
-    with _open_as_binary(content) as data:
-        data_as_bytes = data if isinstance(data, bytes) else data.read()
-        return base64.b64encode(data_as_bytes).decode()
+    raw_bytes = _open_as_mime_bytes(content)
+    return base64.b64encode(raw_bytes).decode()
 
 
 def _as_url(content: ContentT, default_mime_type: str) -> str:
-    if isinstance(content, str) and (content.startswith("https://") or content.startswith("http://")):
+    if isinstance(content, str) and content.startswith(("http://", "https://", "data:")):
         return content
 
-    # Handle MIME type detection for different content types
-    mime_type = None
-    if isinstance(content, (str, Path)):
-        mime_type = mimetypes.guess_type(content, strict=False)[0]
-    elif is_pillow_available():
-        from PIL import Image
+    # Convert content to bytes
+    raw_bytes = _open_as_mime_bytes(content)
 
-        if isinstance(content, Image.Image):
-            # Determine MIME type from PIL Image format, in sync with `_open_as_binary`
-            mime_type = f"image/{(content.format or 'PNG').lower()}"
+    # Get MIME type
+    mime_type = raw_bytes.mime_type or default_mime_type
 
-    mime_type = mime_type or default_mime_type
-    encoded_data = _b64_encode(content)
+    # Encode content to base64
+    encoded_data = base64.b64encode(raw_bytes).decode()
+
+    # Build data URL
     return f"data:{mime_type};base64,{encoded_data}"
 
 
@@ -270,9 +275,6 @@ def _as_dict(response: Union[bytes, Dict]) -> Dict:
     return json.loads(response) if isinstance(response, bytes) else response
 
 
-## PAYLOAD UTILS
-
-
 ## STREAMING UTILS