Merge branch 'main' into v0.11-release

Author: Wauplin

.github/ISSUE_TEMPLATE/bug-report.yml

@@ -36,7 +36,7 @@ body:
         huggingface-cli env
         ```
 
-        If your are working in a notebook, please run it in a code cell:
+        If you are working in a notebook, please run it in a code cell:
         ```py
         from huggingface_hub import dump_environment_info
 

.github/workflows/python-tests.yml

@@ -75,11 +75,11 @@ jobs:
           case "${{ matrix.test_name }}" in
 
             "Repository only")
-              eval "$PYTEST ../tests -k 'RepositoryTest'"
+              eval "$PYTEST ../tests -k 'RepositoryTest or RepositoryDatasetTest'"
               ;;
 
             "Everything else")
-              eval "$PYTEST ../tests -k 'not RepositoryTest'"
+              eval "$PYTEST ../tests -k 'not RepositoryTest and not RepositoryDatasetTest'"
               ;;
 
             lfs)

src/huggingface_hub/_commit_api.py

@@ -471,7 +471,7 @@ def fetch_upload_modes(
             if not path.endswith(".gitkeep"):
                 warnings.warn(
                     f"About to commit an empty file: '{path}'. Are you sure this is"
-                    " intended ?"
+                    " intended?"
                 )
             upload_modes[path] = "regular"
 

src/huggingface_hub/constants.py

@@ -46,6 +46,7 @@ def _is_true_or_auto(value: Optional[str]) -> bool:
 HUGGINGFACE_CO_URL_TEMPLATE = ENDPOINT + "/{repo_id}/resolve/{revision}/{filename}"
 HUGGINGFACE_HEADER_X_REPO_COMMIT = "X-Repo-Commit"
 HUGGINGFACE_HEADER_X_LINKED_ETAG = "X-Linked-Etag"
+HUGGINGFACE_HEADER_X_LINKED_SIZE = "X-Linked-Size"
 
 REPO_ID_SEPARATOR = "--"
 # ^ this substring is not allowed in repo_ids on hf.co

src/huggingface_hub/file_download.py

@@ -26,6 +26,7 @@
     HF_HUB_DISABLE_SYMLINKS_WARNING,
     HUGGINGFACE_CO_URL_TEMPLATE,
     HUGGINGFACE_HEADER_X_LINKED_ETAG,
+    HUGGINGFACE_HEADER_X_LINKED_SIZE,
     HUGGINGFACE_HEADER_X_REPO_COMMIT,
     HUGGINGFACE_HUB_CACHE,
     REPO_ID_SEPARATOR,
@@ -146,7 +147,8 @@ class HfFileMetadata:
         location (`str`):
             Location where to download the file. Can be a Hub url or not (CDN).
         size (`size`):
-            Size of the file.
+            Size of the file. In case of an LFS file, contains the size of the actual
+            LFS file, not the pointer.
     """
 
     commit_hash: Optional[str]
@@ -1384,7 +1386,10 @@ def get_hf_file_metadata(
         # Do not use directly `url`, as `_request_wrapper` might have followed relative
         # redirects.
         location=r.headers.get("Location") or r.request.url,  # type: ignore
-        size=_int_or_none(r.headers.get("Content-Length")),
+        size=_int_or_none(
+            r.headers.get(HUGGINGFACE_HEADER_X_LINKED_SIZE)
+            or r.headers.get("Content-Length")
+        ),
     )
 
 

src/huggingface_hub/repocard.py

@@ -788,47 +788,38 @@ def metadata_update(
             else:
                 existing_results = card.data.eval_results
 
+                # Iterate over new results
+                #   Iterate over existing results
+                #       If both results describe the same metric but value is different:
+                #           If overwrite=True: overwrite the metric value
+                #           Else: raise ValueError
+                #       Else: append new result to existing ones.
                 for new_result in new_results:
                     result_found = False
-                    for existing_result_index, existing_result in enumerate(
-                        existing_results
-                    ):
-                        if all(
-                            [
-                                new_result.dataset_name == existing_result.dataset_name,
-                                new_result.dataset_type == existing_result.dataset_type,
-                                new_result.task_type == existing_result.task_type,
-                                new_result.task_name == existing_result.task_name,
-                                new_result.metric_name == existing_result.metric_name,
-                                new_result.metric_type == existing_result.metric_type,
-                            ]
-                        ):
-                            if (
-                                new_result.metric_value != existing_result.metric_value
-                                and not overwrite
-                            ):
-                                existing_str = (
-                                    f"name: {new_result.metric_name}, type:"
-                                    f" {new_result.metric_type}"
-                                )
+                    for existing_result in existing_results:
+                        if new_result.is_equal_except_value(existing_result):
+                            if new_result != existing_result and not overwrite:
                                 raise ValueError(
                                     "You passed a new value for the existing metric"
-                                    f" '{existing_str}'. Set `overwrite=True` to"
-                                    " overwrite existing metrics."
+                                    f" 'name: {new_result.metric_name}, type: "
+                                    f"{new_result.metric_type}'. Set `overwrite=True`"
+                                    " to overwrite existing metrics."
                                 )
                             result_found = True
-                            card.data.eval_results[existing_result_index] = new_result
+                            existing_result.metric_value = new_result.metric_value
                     if not result_found:
                         card.data.eval_results.append(new_result)
         else:
+            # Any metadata that is not a result metric
             if (
                 hasattr(card.data, key)
                 and getattr(card.data, key) is not None
                 and not overwrite
                 and getattr(card.data, key) != value
             ):
                 raise ValueError(
-                    f"""You passed a new value for the existing meta data field '{key}'. Set `overwrite=True` to overwrite existing metadata."""
+                    f"You passed a new value for the existing meta data field '{key}'."
+                    " Set `overwrite=True` to overwrite existing metadata."
                 )
             else:
                 setattr(card.data, key, value)

src/huggingface_hub/repocard_data.py

@@ -121,6 +121,18 @@ class EvalResult:
     # A JSON Web Token that is used to verify whether the metrics originate from Hugging Face's [evaluation service](https://huggingface.co/spaces/autoevaluate/model-evaluator) or not.
     verify_token: Optional[str] = None
 
+    def is_equal_except_value(self, other: "EvalResult") -> bool:
+        """
+        Return True if `self` and `other` describe exactly the same metric but with a
+        different value.
+        """
+        for key, _ in self.__dict__.items():
+            if key == "metric_value":
+                continue
+            if getattr(self, key) != getattr(other, key):
+                return False
+        return True
+
 
 @dataclass
 class CardData:

src/huggingface_hub/repository.py

@@ -5,6 +5,7 @@
 import tempfile
 import threading
 import time
+import warnings
 from contextlib import contextmanager
 from pathlib import Path
 from typing import Callable, Dict, Iterator, List, Optional, Tuple, Union
@@ -15,7 +16,14 @@
 
 from .hf_api import HfApi, repo_type_and_id_from_hf_id
 from .lfs import LFS_MULTIPART_UPLOAD_COMMAND
-from .utils import HfFolder, logging, run_subprocess, tqdm, validate_hf_hub_args
+from .utils import (
+    HfFolder,
+    RepositoryNotFoundError,
+    logging,
+    run_subprocess,
+    tqdm,
+    validate_hf_hub_args,
+)
 from .utils._deprecation import _deprecate_arguments, _deprecate_method
 from .utils._typing import TypedDict
 
@@ -684,9 +692,10 @@ def clone_from(self, repo_url: str, token: Union[bool, str, None] = None):
         if hub_url in repo_url or (
             "http" not in repo_url and len(repo_url.split("/")) <= 2
         ):
-            repo_type, namespace, repo_id = repo_type_and_id_from_hf_id(
+            repo_type, namespace, repo_name = repo_type_and_id_from_hf_id(
                 repo_url, hub_url=hub_url
             )
+            repo_id = f"{namespace}/{repo_name}" if namespace is not None else repo_name
 
             if repo_type is not None:
                 self._repo_type = repo_type
@@ -701,10 +710,32 @@ def clone_from(self, repo_url: str, token: Union[bool, str, None] = None):
                 scheme = urlparse(repo_url).scheme
                 repo_url = repo_url.replace(f"{scheme}://", f"{scheme}://user:{token}@")
 
-            if namespace is not None:
-                repo_url += f"{namespace}/"
             repo_url += repo_id
 
+            # To be removed: check if repo exists. If not, create it first.
+            try:
+                HfApi().repo_info(f"{repo_id}", repo_type=self._repo_type, token=token)
+            except RepositoryNotFoundError:
+                if self._repo_type == "space":
+                    raise ValueError(
+                        "Creating a Space through passing Space link to clone_from is"
+                        " not allowed. Make sure the Space exists on Hugging Face Hub."
+                    )
+                else:
+                    warnings.warn(
+                        "Creating a repository through 'clone_from' is deprecated and"
+                        " will be removed in v0.12. Please create the repository first"
+                        " using `create_repo(..., exists_ok=True)`.",
+                        FutureWarning,
+                    )
+                    self.client.create_repo(
+                        repo_id=repo_id,
+                        token=token,
+                        repo_type=self._repo_type,
+                        exist_ok=True,
+                        private=self._private,
+                    )
+
         # For error messages, it's cleaner to show the repo url without the token.
         clean_repo_url = re.sub(r"(https?)://.*@", r"\1://", repo_url)
         try:

src/huggingface_hub/utils/_runtime.py

@@ -182,6 +182,13 @@ def is_google_colab() -> bool:
 
 
 def dump_environment_info() -> Dict[str, Any]:
+    """Dump information about the machine to help debugging issues.
+
+    Similar helper exist in:
+    - `datasets` (https://github.com/huggingface/datasets/blob/main/src/datasets/commands/env.py)
+    - `diffusers` (https://github.com/huggingface/diffusers/blob/main/src/diffusers/commands/env.py)
+    - `transformers` (https://github.com/huggingface/transformers/blob/main/src/transformers/commands/env.py)
+    """
     from huggingface_hub import HfFolder, whoami
     from huggingface_hub.utils import list_credential_helpers
 

tests/test_file_download.py

@@ -376,6 +376,17 @@ def test_get_hf_file_metadata_from_a_renamed_repo(self) -> None:
             url.replace(DUMMY_RENAMED_OLD_MODEL_ID, DUMMY_RENAMED_NEW_MODEL_ID),
         )
 
+    def test_get_hf_file_metadata_from_a_lfs_file(self) -> None:
+        """Test getting metadata from an LFS file.
+
+        Must get size of the LFS file, not size of the pointer file
+        """
+        url = hf_hub_url("gpt2", filename="tf_model.h5")
+        metadata = get_hf_file_metadata(url)
+
+        self.assertIn("cdn-lfs", metadata.location)  # Redirection
+        self.assertEqual(metadata.size, 497933648)  # Size of LFS file, not pointer
+
 
 class StagingCachedDownloadTest(unittest.TestCase):
     def test_download_from_a_gated_repo_with_hf_hub_download(self):