trigger CI for #40889 (#41070)

* Adapt and test huggingface_hub v1.0.0.rc0

* forgot to bump hfh

* bump

* code quality

* code quality

* relax dependency table

* fix has_file

* install hfh 1.0.0.rc0 in circle ci jobs

* repostiryo

* push to hub now returns a commit url

* catch HfHubHTTPError

* check commit on branch

* add it back

* fix ?

* remove deprecated test

* uncomment another test

* trigger

* no proxies

* many more small changes

* fix load PIL Image from httpx

* require 1.0.0.rc0

* fix mocked tests

* fix others

* unchange

* unchange

* args

* Update .circleci/config.yml

---------

Co-authored-by: Wauplin <lucainp@gmail.com>

Author: ydshieh

setup.py

@@ -114,7 +114,7 @@
     "GitPython<3.1.19",
     "hf-doc-builder>=0.3.0",
     "hf_xet",
-    "huggingface-hub>=0.34.0,<1.0",
+    "huggingface-hub==1.0.0.rc0",
     "importlib_metadata",
     "ipadic>=1.0.0,<2.0",
     "jinja2>=3.1.0",

src/transformers/audio_utils.py

@@ -25,8 +25,8 @@
 from io import BytesIO
 from typing import Any, Optional, Union
 
+import httpx
 import numpy as np
-import requests
 from packaging import version
 
 from .utils import (
@@ -131,7 +131,9 @@ def load_audio_librosa(audio: Union[str, np.ndarray], sampling_rate=16000, timeo
 
     # Load audio from URL (e.g https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen2-Audio/audio/translate_to_chinese.wav)
     if audio.startswith("http://") or audio.startswith("https://"):
-        audio = librosa.load(BytesIO(requests.get(audio, timeout=timeout).content), sr=sampling_rate)[0]
+        audio = librosa.load(
+            BytesIO(httpx.get(audio, follow_redirects=True, timeout=timeout).content), sr=sampling_rate
+        )[0]
     elif os.path.isfile(audio):
         audio = librosa.load(audio, sr=sampling_rate)[0]
     return audio
@@ -173,7 +175,7 @@ def load_audio_as(
         # Load audio bytes from URL or file
         audio_bytes = None
         if audio.startswith(("http://", "https://")):
-            response = requests.get(audio, timeout=timeout)
+            response = httpx.get(audio, follow_redirects=True, timeout=timeout)
             response.raise_for_status()
             audio_bytes = response.content
         elif os.path.isfile(audio):

src/transformers/dependency_versions_table.py

@@ -23,7 +23,7 @@
     "GitPython": "GitPython<3.1.19",
     "hf-doc-builder": "hf-doc-builder>=0.3.0",
     "hf_xet": "hf_xet",
-    "huggingface-hub": "huggingface-hub>=0.34.0,<1.0",
+    "huggingface-hub": "huggingface-hub==1.0.0.rc0",
     "importlib_metadata": "importlib_metadata",
     "ipadic": "ipadic>=1.0.0,<2.0",
     "jinja2": "jinja2>=3.1.0",

src/transformers/file_utils.py

@@ -17,9 +17,6 @@
 This module should not be update anymore and is only left for backward compatibility.
 """
 
-from huggingface_hub import get_full_repo_name  # for backward compatibility
-from huggingface_hub.constants import HF_HUB_DISABLE_TELEMETRY as DISABLE_TELEMETRY  # for backward compatibility
-
 from . import __version__
 
 # Backward compatibility imports, to make sure all those objects can be found in file_utils

src/transformers/image_utils.py

@@ -19,8 +19,8 @@
 from io import BytesIO
 from typing import Optional, Union
 
+import httpx
 import numpy as np
-import requests
 
 from .utils import (
     ExplicitEnum,
@@ -462,7 +462,7 @@ def load_image(image: Union[str, "PIL.Image.Image"], timeout: Optional[float] =
         if image.startswith("http://") or image.startswith("https://"):
             # We need to actually check for a real protocol, otherwise it's impossible to use a local file
             # like http_huggingface_co.png
-            image = PIL.Image.open(BytesIO(requests.get(image, timeout=timeout).content))
+            image = PIL.Image.open(BytesIO(httpx.get(image, timeout=timeout, follow_redirects=True).content))
         elif os.path.isfile(image):
             image = PIL.Image.open(image)
         else:

src/transformers/modelcard.py

@@ -21,7 +21,7 @@
 from pathlib import Path
 from typing import Any, Optional, Union
 
-import requests
+import httpx
 import yaml
 from huggingface_hub import model_info
 from huggingface_hub.errors import OfflineModeIsEnabled
@@ -380,12 +380,7 @@ def __post_init__(self):
                 for tag in info.tags:
                     if tag.startswith("license:"):
                         self.license = tag[8:]
-            except (
-                requests.exceptions.HTTPError,
-                requests.exceptions.ConnectionError,
-                HFValidationError,
-                OfflineModeIsEnabled,
-            ):
+            except (httpx.HTTPError, HFValidationError, OfflineModeIsEnabled):
                 pass
 
     def create_model_index(self, metric_mapping):

src/transformers/pipelines/audio_classification.py

@@ -14,8 +14,8 @@
 import subprocess
 from typing import Any, Union
 
+import httpx
 import numpy as np
-import requests
 
 from ..utils import add_end_docstrings, is_torch_available, is_torchaudio_available, is_torchcodec_available, logging
 from .base import Pipeline, build_pipeline_init_args
@@ -168,7 +168,7 @@ def preprocess(self, inputs):
             if inputs.startswith("http://") or inputs.startswith("https://"):
                 # We need to actually check for a real protocol, otherwise it's impossible to use a local file
                 # like http_huggingface_co.png
-                inputs = requests.get(inputs).content
+                inputs = httpx.get(inputs, follow_redirects=True).content
             else:
                 with open(inputs, "rb") as f:
                     inputs = f.read()

src/transformers/pipelines/automatic_speech_recognition.py

@@ -14,8 +14,8 @@
 from collections import defaultdict
 from typing import TYPE_CHECKING, Any, Optional, Union
 
+import httpx
 import numpy as np
-import requests
 
 from ..generation import GenerationConfig
 from ..tokenization_utils import PreTrainedTokenizer
@@ -355,7 +355,7 @@ def preprocess(self, inputs, chunk_length_s=0, stride_length_s=None):
             if inputs.startswith("http://") or inputs.startswith("https://"):
                 # We need to actually check for a real protocol, otherwise it's impossible to use a local file
                 # like http_huggingface_co.png
-                inputs = requests.get(inputs).content
+                inputs = httpx.get(inputs, follow_redirects=True).content
             else:
                 with open(inputs, "rb") as f:
                     inputs = f.read()

src/transformers/pipelines/image_to_image.py

@@ -46,12 +46,13 @@ class ImageToImagePipeline(Pipeline):
 
     ```python
     >>> from PIL import Image
-    >>> import requests
+    >>> import httpx
+    >>> import io
 
     >>> from transformers import pipeline
 
     >>> upscaler = pipeline("image-to-image", model="caidas/swin2SR-classical-sr-x2-64")
-    >>> img = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+    >>> img = Image.open(io.BytesIO(httpx.get("http://images.cocodataset.org/val2017/000000039769.jpg").content))
     >>> img = img.resize((64, 64))
     >>> upscaled_img = upscaler(img)
     >>> img.size

src/transformers/pipelines/video_classification.py

@@ -15,7 +15,7 @@
 from io import BytesIO
 from typing import Any, Optional, Union, overload
 
-import requests
+import httpx
 
 from ..utils import (
     add_end_docstrings,
@@ -142,7 +142,7 @@ def preprocess(self, video, num_frames=None, frame_sampling_rate=1):
             num_frames = self.model.config.num_frames
 
         if video.startswith("http://") or video.startswith("https://"):
-            video = BytesIO(requests.get(video).content)
+            video = BytesIO(httpx.get(video, follow_redirects=True).content)
 
         container = av.open(video)