Merge branch 'main' into verify-checksum-cli

hanouticelina · web-flow · commit aef6bd8902e7 · 2025-10-31T12:14:19.000+01:00
diff --git a/.github/workflows/python-tests.yml b/.github/workflows/python-tests.yml
@@ -21,10 +21,10 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.9", "3.13"]
+        python-version: ["3.9", "3.14"]
         test_name: ["Everything else", "Inference only", "Xet only"]
         include:
-          - python-version: "3.13" # LFS not ran on 3.9
+          - python-version: "3.14" # LFS not ran on 3.9
             test_name: "lfs"
           - python-version: "3.9"
             test_name: "fastai"
@@ -34,6 +34,8 @@ jobs:
             test_name: "Python 3.9, torch_1.11"
           - python-version: "3.12" # test torch latest on python 3.12 only.
             test_name: "torch_latest"
+          - python-version: "3.13" # gradio not supported on 3.14 -> test it on 3.13
+            test_name: "gradio"
     steps:
       - uses: actions/checkout@v2
       - name: Set up Python ${{ matrix.python-version }}
@@ -69,6 +71,10 @@ jobs:
               uv pip install "huggingface_hub[fastai] @ ."
               ;;
 
+            gradio)
+              uv pip install "huggingface_hub[gradio] @ ."
+            ;;
+
             torch_latest)
               uv pip install "huggingface_hub[torch] @ ."
               uv pip install --upgrade torch
@@ -117,6 +123,10 @@ jobs:
               eval "$PYTEST ../tests/test_fastai*"
             ;;
 
+            gradio)
+              eval "$PYTEST ../tests/test_webhooks_server.py"
+            ;;
+
             "Python 3.9, torch_1.11" | torch_latest)
             eval "$PYTEST ../tests/test_hub_mixin*"
             eval "$PYTEST ../tests/test_serialization.py"
@@ -148,7 +158,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.9", "3.11"]
+        python-version: ["3.9", "3.14"]
         test_name: ["Everything else", "Xet only"]
 
     steps:
diff --git a/docs/source/en/guides/cli.md b/docs/source/en/guides/cli.md
@@ -4,7 +4,7 @@ rendered properly in your Markdown viewer.
 
 # Command Line Interface (CLI)
 
-The `huggingface_hub` Python package comes with a built-in CLI called `hf`. This tool allows you to interact with the Hugging Face Hub directly from a terminal. For example, you can login to your account, create a repository, upload and download files, etc. It also comes with handy features to configure your machine or manage your cache. In this guide, we will have a look at the main features of the CLI and how to use them.
+The `huggingface_hub` Python package comes with a built-in CLI called `hf`. This tool allows you to interact with the Hugging Face Hub directly from a terminal. For example, you can log in to your account, create a repository, upload and download files, etc. It also comes with handy features to configure your machine or manage your cache. In this guide, we will have a look at the main features of the CLI and how to use them.
 
 > [!TIP]
 > This guide covers the most important features of the `hf` CLI.
@@ -174,7 +174,7 @@ hf download --help
 
 ### Download a single file
 
-To download a single file from a repo, simply provide the repo_id and filename as follow:
+To download a single file from a repo, simply provide the repo_id and filename as follows:
 
 ```bash
 >>> hf download gpt2 config.json
diff --git a/docs/source/en/guides/download.md b/docs/source/en/guides/download.md
@@ -68,7 +68,7 @@ Note that it is used internally by [`hf_hub_download`].
 ## Download an entire repository
 
 [`snapshot_download`] downloads an entire repository at a given revision. It uses internally [`hf_hub_download`] which
-means all downloaded files are also cached on your local disk. Downloads are made concurrently to speed-up the process.
+means all downloaded files are also cached on your local disk. Downloads are made concurrently to speed up the process.
 
 To download a whole repository, just pass the `repo_id` and `repo_type`:
 
diff --git a/docs/source/en/guides/jobs.md b/docs/source/en/guides/jobs.md
@@ -172,7 +172,7 @@ Running this will show the following output!
 This code ran with the following GPU: NVIDIA A10G
 ```
 
-Use this to run a fine tuning script like [trl/scripts/sft.py](https://github.com/huggingface/trl/blob/main/trl/scripts/sft.py) with UV:
+Use this to run a fine-tuning script like [trl/scripts/sft.py](https://github.com/huggingface/trl/blob/main/trl/scripts/sft.py) with UV:
 
 ```python
 >>> from huggingface_hub import run_uv_job
diff --git a/docs/source/en/guides/manage-spaces.md b/docs/source/en/guides/manage-spaces.md
@@ -10,7 +10,7 @@ In this guide, we will see how to manage your Space runtime
 
 ## A simple example: configure secrets and hardware.
 
-Here is an end-to-end example to create and setup a Space on the Hub.
+Here is an end-to-end example to create and set up a Space on the Hub.
 
 **1. Create a Space on the Hub.**
 
diff --git a/docs/source/en/package_reference/environment_variables.md b/docs/source/en/package_reference/environment_variables.md
@@ -125,7 +125,7 @@ If `HF_HUB_OFFLINE=1` is set as environment variable and you call any method of
 
 ### HF_HUB_DISABLE_IMPLICIT_TOKEN
 
-Authentication is not mandatory for every requests to the Hub. For instance, requesting
+Authentication is not mandatory for every request to the Hub. For instance, requesting
 details about `"gpt2"` model does not require to be authenticated. However, if a user is
 [logged in](../package_reference/login), the default behavior will be to always send the token
 in order to ease user experience (never get a HTTP 401 Unauthorized) when accessing private or gated repositories. For privacy, you can
@@ -138,7 +138,7 @@ would need to explicitly pass `token=True` argument in your script.
 
 ### HF_HUB_DISABLE_PROGRESS_BARS
 
-For time consuming tasks, `huggingface_hub` displays a progress bar by default (using tqdm).
+For time-consuming tasks, `huggingface_hub` displays a progress bar by default (using tqdm).
 You can disable all the progress bars at once by setting `HF_HUB_DISABLE_PROGRESS_BARS=1`.
 
 ### HF_HUB_DISABLE_SYMLINKS_WARNING
diff --git a/setup.py b/setup.py
@@ -65,7 +65,6 @@ def get_version() -> str:
         "urllib3<2.0",  # VCR.py broken with urllib3 2.0 (see https://urllib3.readthedocs.io/en/stable/v2-migration-guide.html)
         "soundfile",
         "Pillow",
-        "requests",  # for gradio
         "numpy",  # for embeddings
         "fastapi",  # To build the documentation
     ]
@@ -74,8 +73,10 @@ def get_version() -> str:
 if sys.version_info >= (3, 10):
     # We need gradio to test webhooks server
     # But gradio 5.0+ only supports python 3.10+ so we don't want to test earlier versions
-    extras["testing"].append("gradio>=5.0.0")
-    extras["testing"].append("requests")  # see https://github.com/gradio-app/gradio/pull/11830
+    extras["gradio"] = [
+        "gradio>=5.0.0",
+        "requests",  # see https://github.com/gradio-app/gradio/pull/11830
+    ]
 
 # Typing extra dependencies list is duplicated in `.pre-commit-config.yaml`
 # Please make sure to update the list there when adding a new typing dependency.
@@ -135,6 +136,7 @@ def get_version() -> str:
         "Programming Language :: Python :: 3.11",
         "Programming Language :: Python :: 3.12",
         "Programming Language :: Python :: 3.13",
+        "Programming Language :: Python :: 3.14",
         "Topic :: Scientific/Engineering :: Artificial Intelligence",
     ],
     include_package_data=True,
diff --git a/src/huggingface_hub/_inference_endpoints.py b/src/huggingface_hub/_inference_endpoints.py
@@ -368,7 +368,7 @@ def scale_to_zero(self) -> "InferenceEndpoint":
         """Scale Inference Endpoint to zero.
 
         An Inference Endpoint scaled to zero will not be charged. It will be resumed on the next request to it, with a
-        cold start delay. This is different than pausing the Inference Endpoint with [`InferenceEndpoint.pause`], which
+        cold start delay. This is different from pausing the Inference Endpoint with [`InferenceEndpoint.pause`], which
         would require a manual resume with [`InferenceEndpoint.resume`].
 
         This is an alias for [`HfApi.scale_to_zero_inference_endpoint`]. The current object is mutated in place with the
diff --git a/src/huggingface_hub/_login.py b/src/huggingface_hub/_login.py
@@ -123,7 +123,7 @@ def logout(token_name: Optional[str] = None) -> None:
 
     Args:
         token_name (`str`, *optional*):
-            Name of the access token to logout from. If `None`, will logout from all saved access tokens.
+            Name of the access token to logout from. If `None`, will log out from all saved access tokens.
     Raises:
         [`ValueError`](https://docs.python.org/3/library/exceptions.html#ValueError):
             If the access token name is not found.
diff --git a/src/huggingface_hub/_space_api.py b/src/huggingface_hub/_space_api.py
@@ -111,7 +111,7 @@ class SpaceRuntime:
             Current hardware of the space. Example: "cpu-basic". Can be `None` if Space
             is `BUILDING` for the first time.
         requested_hardware (`str` or `None`):
-            Requested hardware. Can be different than `hardware` especially if the request
+            Requested hardware. Can be different from `hardware` especially if the request
             has just been made. Example: "t4-medium". Can be `None` if no hardware has
             been requested yet.
         sleep_time (`int` or `None`):
diff --git a/src/huggingface_hub/_webhooks_server.py b/src/huggingface_hub/_webhooks_server.py
@@ -50,7 +50,7 @@ class WebhooksServer:
     It is recommended to accept [`WebhookPayload`] as the first argument of the webhook function. It is a Pydantic
     model that contains all the information about the webhook event. The data will be parsed automatically for you.
 
-    Check out the [webhooks guide](../guides/webhooks_server) for a step-by-step tutorial on how to setup your
+    Check out the [webhooks guide](../guides/webhooks_server) for a step-by-step tutorial on how to set up your
     WebhooksServer and deploy it on a Space.
 
     > [!WARNING]
@@ -231,7 +231,7 @@ def webhook_endpoint(path: Optional[str] = None) -> Callable:
     you can use [`WebhooksServer`] directly. You can register multiple webhook endpoints (to the same server) by using
     this decorator multiple times.
 
-    Check out the [webhooks guide](../guides/webhooks_server) for a step-by-step tutorial on how to setup your
+    Check out the [webhooks guide](../guides/webhooks_server) for a step-by-step tutorial on how to set up your
     server and deploy it on a Space.
 
     > [!WARNING]
diff --git a/src/huggingface_hub/constants.py b/src/huggingface_hub/constants.py
@@ -233,7 +233,7 @@ def _as_int(value: Optional[str]) -> Optional[int]:
 # Used to override the get request timeout on a system level
 HF_HUB_DOWNLOAD_TIMEOUT: int = _as_int(os.environ.get("HF_HUB_DOWNLOAD_TIMEOUT")) or DEFAULT_DOWNLOAD_TIMEOUT
 
-# Allows to add information about the requester in the user-agent (eg. partner name)
+# Allows to add information about the requester in the user-agent (e.g. partner name)
 HF_HUB_USER_AGENT_ORIGIN: Optional[str] = os.environ.get("HF_HUB_USER_AGENT_ORIGIN")
 
 # If OAuth didn't work after 2 redirects, there's likely a third-party cookie issue in the Space iframe view.
diff --git a/src/huggingface_hub/dataclasses.py b/src/huggingface_hub/dataclasses.py
@@ -305,12 +305,7 @@ def validate_typed_dict(schema: type[TypedDictType], data: dict) -> None:
 @lru_cache
 def _build_strict_cls_from_typed_dict(schema: type[TypedDictType]) -> Type:
     # Extract type hints from the TypedDict class
-    type_hints = {
-        # We do not use `get_type_hints` here to avoid evaluating ForwardRefs (which might fail).
-        # ForwardRefs are not validated by @strict anyway.
-        name: value if value is not None else type(None)
-        for name, value in schema.__dict__.get("__annotations__", {}).items()
-    }
+    type_hints = _get_typed_dict_annotations(schema)
 
     # If the TypedDict is not total, wrap fields as NotRequired (unless explicitly Required or NotRequired)
     if not getattr(schema, "__total__", True):
@@ -338,6 +333,22 @@ def _build_strict_cls_from_typed_dict(schema: type[TypedDictType]) -> Type:
     return strict(make_dataclass(schema.__name__, fields))
 
 
+def _get_typed_dict_annotations(schema: type[TypedDictType]) -> dict[str, Any]:
+    """Extract type annotations from a TypedDict class."""
+    try:
+        # Available in Python 3.14+
+        import annotationlib
+
+        return annotationlib.get_annotations(schema)
+    except ImportError:
+        return {
+            # We do not use `get_type_hints` here to avoid evaluating ForwardRefs (which might fail).
+            # ForwardRefs are not validated by @strict anyway.
+            name: value if value is not None else type(None)
+            for name, value in schema.__dict__.get("__annotations__", {}).items()
+        }
+
+
 def validated_field(
     validator: Union[list[Validator_T], Validator_T],
     default: Union[Any, _MISSING_TYPE] = MISSING,
diff --git a/src/huggingface_hub/file_download.py b/src/huggingface_hub/file_download.py
@@ -1034,7 +1034,7 @@ def _hf_hub_download_to_cache_dir(
     locks_dir = os.path.join(cache_dir, ".locks")
     storage_folder = os.path.join(cache_dir, repo_folder_name(repo_id=repo_id, repo_type=repo_type))
 
-    # cross platform transcription of filename, to be used as a local file path.
+    # cross-platform transcription of filename, to be used as a local file path.
     relative_filename = os.path.join(*filename.split("/"))
     if os.name == "nt":
         if relative_filename.startswith("..\\") or "\\..\\" in relative_filename:
@@ -1531,7 +1531,7 @@ def get_hf_file_metadata(
     # Return
     return HfFileMetadata(
         commit_hash=response.headers.get(constants.HUGGINGFACE_HEADER_X_REPO_COMMIT),
-        # We favor a custom header indicating the etag of the linked resource, and we fallback to the regular etag header.
+        # We favor a custom header indicating the etag of the linked resource, and we fall back to the regular etag header.
         etag=_normalize_etag(
             response.headers.get(constants.HUGGINGFACE_HEADER_X_LINKED_ETAG) or response.headers.get("ETag")
         ),
diff --git a/src/huggingface_hub/hub_mixin.py b/src/huggingface_hub/hub_mixin.py
@@ -117,7 +117,7 @@ class ModelHubMixin:
             Tags to be added to the model card. Used to generate model card. E.g. ["computer-vision"]
         coders (`dict[Type, tuple[Callable, Callable]]`, *optional*):
             Dictionary of custom types and their encoders/decoders. Used to encode/decode arguments that are not
-            jsonable by default. E.g dataclasses, argparse.Namespace, OmegaConf, etc.
+            jsonable by default. E.g. dataclasses, argparse.Namespace, OmegaConf, etc.
 
     Example:
 
diff --git a/src/huggingface_hub/serialization/_dduf.py b/src/huggingface_hub/serialization/_dduf.py
@@ -202,7 +202,7 @@ def export_entries_as_dduf(
         ... # ... do some work with the pipeline
 
         >>> def as_entries(pipe: DiffusionPipeline) -> Generator[tuple[str, bytes], None, None]:
-        ...     # Build an generator that yields the entries to add to the DDUF file.
+        ...     # Build a generator that yields the entries to add to the DDUF file.
         ...     # The first element of the tuple is the filename in the DDUF archive (must use UNIX separator!). The second element is the content of the file.
         ...     # Entries will be evaluated lazily when the DDUF file is created (only 1 entry is loaded in memory at a time)
         ...     yield "vae/config.json", pipe.vae.to_json_string().encode()
diff --git a/src/huggingface_hub/serialization/_torch.py b/src/huggingface_hub/serialization/_torch.py
@@ -710,7 +710,7 @@ def _get_unique_id(tensor: "torch.Tensor") -> Union[int, tuple[Any, ...]]:
         pass
 
     if tensor.device.type == "xla" and is_torch_tpu_available():
-        # NOTE: xla tensors dont have storage
+        # NOTE: xla tensors don't have storage
         # use some other unique id to distinguish.
         # this is a XLA tensor, it must be created using torch_xla's
         # device. So the following import is safe:
@@ -761,7 +761,7 @@ def get_torch_storage_size(tensor: "torch.Tensor") -> int:
             attrs, _ = tensor.__tensor_flatten__()  # type: ignore[attr-defined]
             return sum(get_torch_storage_size(getattr(tensor, attr)) for attr in attrs)
     except ImportError:
-        # for torch version less than 2.1, we can fallback to original implementation
+        # for torch version less than 2.1, we can fall back to original implementation
         pass
 
     try:
@@ -808,7 +808,7 @@ def storage_ptr(tensor: "torch.Tensor") -> Union[int, tuple[Any, ...]]:
         if is_traceable_wrapper_subclass(tensor):
             return _get_unique_id(tensor)  # type: ignore
     except ImportError:
-        # for torch version less than 2.1, we can fallback to original implementation
+        # for torch version less than 2.1, we can fall back to original implementation
         pass
 
     try:
@@ -916,7 +916,7 @@ def _is_complete(tensor: "torch.Tensor") -> bool:
             attrs, _ = tensor.__tensor_flatten__()  # type: ignore[attr-defined]
             return all(_is_complete(getattr(tensor, attr)) for attr in attrs)
     except ImportError:
-        # for torch version less than 2.1, we can fallback to original implementation
+        # for torch version less than 2.1, we can fall back to original implementation
         pass
 
     return tensor.data_ptr() == storage_ptr(tensor) and tensor.nelement() * _get_dtype_size(
diff --git a/src/huggingface_hub/utils/_git_credential.py b/src/huggingface_hub/utils/_git_credential.py
@@ -111,7 +111,7 @@ def _parse_credential_output(output: str) -> list[str]:
         output (`str`):
             The output of `git credential fill`.
     """
-    # NOTE: If user has set an helper for a custom URL, it will not be caught here.
+    # NOTE: If user has set a helper for a custom URL, it will not be caught here.
     #       Example: `credential.https://huggingface.co.helper=store`
     #       See: https://github.com/huggingface/huggingface_hub/pull/1138#discussion_r1013324508
     return sorted(  # Sort for nice printing
diff --git a/src/huggingface_hub/utils/_http.py b/src/huggingface_hub/utils/_http.py
@@ -17,6 +17,7 @@
 import atexit
 import io
 import json
+import os
 import re
 import threading
 import time
@@ -244,6 +245,8 @@ def close_session() -> None:
 
 
 atexit.register(close_session)
+if hasattr(os, "register_at_fork"):
+    os.register_at_fork(after_in_child=close_session)
 
 
 def _http_backoff_base(
diff --git a/src/huggingface_hub/utils/_telemetry.py b/src/huggingface_hub/utils/_telemetry.py
@@ -25,7 +25,7 @@ def send_telemetry(
     user_agent: Union[dict, str, None] = None,
 ) -> None:
     """
-    Sends telemetry that helps tracking usage of different HF libraries.
+    Sends telemetry that helps track usage of different HF libraries.
 
     This usage data helps us debug issues and prioritize new features. However, we understand that not everyone wants
     to share additional information, and we respect your privacy. You can disable telemetry collection by setting the
diff --git a/src/huggingface_hub/utils/_validators.py b/src/huggingface_hub/utils/_validators.py
@@ -161,7 +161,7 @@ def smoothly_deprecate_legacy_arguments(fn_name: str, kwargs: dict[str, Any]) ->
             In particular, it is not possible to configure proxies on a per-request basis. The solution is to configure
             it globally using the [`set_client_factory`] function or using the HTTP_PROXY environment variable.
 
-            More more details, see:
+            For more details, see:
             - https://www.python-httpx.org/advanced/proxies/
             - https://www.python-httpx.org/compatibility/#proxy-keys.
 
diff --git a/tests/test_hf_api.py b/tests/test_hf_api.py
@@ -837,7 +837,7 @@ def test_commit_preflight_on_lots_of_lfs_files(self, repo_url: RepoUrl):
     def test_create_commit_repo_id_case_insensitive(self):
         """Test create commit but repo_id is lowercased.
 
-        Regression test for #1371. Hub API is already case insensitive. Somehow the issue was with the `requests`
+        Regression test for #1371. Hub API is already case-insensitive. Somehow the issue was with the `requests`
         streaming implementation when generating the ndjson payload "on the fly". It seems that the server was
         receiving only the first line which causes a confusing "400 Bad Request - Add a line with the key `lfsFile`,
         `file` or `deletedFile`". Passing raw bytes instead of a generator fixes the problem.
@@ -3958,7 +3958,7 @@ def test_create_collection_exists_ok(self) -> None:
 
         self.assertEqual(collection_1.slug, collection_2.slug)
         self.assertIsNone(collection_1.description)
-        self.assertIsNone(collection_2.description)  # Did not got updated!
+        self.assertIsNone(collection_2.description)  # Did not get updated!
 
     def test_create_private_collection(self) -> None:
         collection = self._api.create_collection(self.title, private=True)
diff --git a/tests/test_utils_cache.py b/tests/test_utils_cache.py
diff --git a/tests/test_utils_http.py b/tests/test_utils_http.py
diff --git a/utils/generate_inference_types.py b/utils/generate_inference_types.py
