In a local JupyterLab that is not a Google Colab environment, _get_token_from_google_colab freezes and stops responding

[tanaga9]

Describe the bug

In a local JupyterLab that is not a Google Colab environment, _get_token_from_google_colab freezes and stops responding.

This issue was temporarily resolved by the following code.

import huggingface_hub.utils._token
huggingface_hub.utils._token._get_token_from_google_colab = lambda: False

Reproduction

• In a local JupyterLab that is not a Google Colab environment
• call StableDiffusionControlNetImg2ImgPipeline.from_single_file

I will attach the traceback of an interrupt during a freeze to the logs.

Logs

---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
<ipython-input-7-30de77117569> in <cell line: 1>()
----> 1 pipe = StableDiffusionControlNetImg2ImgPipeline.from_single_file(
      2     #"https://huggingface.co/OedoSoldier/ambientmix",
      3     "ambientmix-half.safetensors",
      4       # "ambientmixAnAnime_v10.safetensors",
      5       #"../stable-diffusion-webui/models/Stable-diffusion/ambientmixAnAnime_v10.safetensors",

/usr/local/lib/python3.10/dist-packages/diffusers/loaders.py in from_single_file(cls, pretrained_model_link_or_path, **kwargs)
   1920             )
   1921 
-> 1922         pipe = download_from_original_stable_diffusion_ckpt(
   1923             pretrained_model_link_or_path,
   1924             pipeline_class=cls,

/usr/local/lib/python3.10/dist-packages/diffusers/pipelines/stable_diffusion/convert_from_ckpt.py in download_from_original_stable_diffusion_ckpt(checkpoint_path, original_config_file, image_size, prediction_type, model_type, extract_ema, scheduler_type, num_in_channels, upcast_attention, device, from_safetensors, stable_unclip, stable_unclip_prior, clip_stats_path, controlnet, load_safety_checker, pipeline_class, local_files_only, vae_path, vae, text_encoder, tokenizer, config_files)
   1532         )
   1533     elif model_type == "FrozenCLIPEmbedder":
-> 1534         text_model = convert_ldm_clip_checkpoint(
   1535             checkpoint, local_files_only=local_files_only, text_encoder=text_encoder
   1536         )

/usr/local/lib/python3.10/dist-packages/diffusers/pipelines/stable_diffusion/convert_from_ckpt.py in convert_ldm_clip_checkpoint(checkpoint, local_files_only, text_encoder)
    779     if text_encoder is None:
    780         config_name = "openai/clip-vit-large-patch14"
--> 781         config = CLIPTextConfig.from_pretrained(config_name, local_files_only=local_files_only)
    782 
    783         ctx = init_empty_weights if is_accelerate_available() else nullcontext

/usr/local/lib/python3.10/dist-packages/transformers/models/clip/configuration_clip.py in from_pretrained(cls, pretrained_model_name_or_path, **kwargs)
    128     @classmethod
    129     def from_pretrained(cls, pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> "PretrainedConfig":
--> 130         config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)
    131 
    132         # get the text config dict if we are loading from CLIPConfig

/usr/local/lib/python3.10/dist-packages/transformers/configuration_utils.py in get_config_dict(cls, pretrained_model_name_or_path, **kwargs)
    572         original_kwargs = copy.deepcopy(kwargs)
    573         # Get config dict associated with the base config file
--> 574         config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs)
    575         if "_commit_hash" in config_dict:
    576             original_kwargs["_commit_hash"] = config_dict["_commit_hash"]

/usr/local/lib/python3.10/dist-packages/transformers/configuration_utils.py in _get_config_dict(cls, pretrained_model_name_or_path, **kwargs)
    627             try:
    628                 # Load from local folder or from cache or download from model Hub and cache
--> 629                 resolved_config_file = cached_file(
    630                     pretrained_model_name_or_path,
    631                     configuration_file,

/usr/local/lib/python3.10/dist-packages/transformers/utils/hub.py in cached_file(path_or_repo_id, filename, cache_dir, force_download, resume_download, proxies, use_auth_token, revision, local_files_only, subfolder, repo_type, user_agent, _raise_exceptions_for_missing_entries, _raise_exceptions_for_connection_errors, _commit_hash)
    415     try:
    416         # Load from URL or cache if already cached
--> 417         resolved_file = hf_hub_download(
    418             path_or_repo_id,
    419             filename,

/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_validators.py in _inner_fn(*args, **kwargs)
    116             kwargs = smoothly_deprecate_use_auth_token(fn_name=fn.__name__, has_token=has_token, kwargs=kwargs)
    117 
--> 118         return fn(*args, **kwargs)
    119 
    120     return _inner_fn  # type: ignore

/usr/local/lib/python3.10/dist-packages/huggingface_hub/file_download.py in hf_hub_download(repo_id, filename, subfolder, repo_type, revision, library_name, library_version, cache_dir, local_dir, local_dir_use_symlinks, user_agent, force_download, force_filename, proxies, etag_timeout, resume_download, token, local_files_only, legacy_cache_layout, endpoint)
   1221     url = hf_hub_url(repo_id, filename, repo_type=repo_type, revision=revision, endpoint=endpoint)
   1222 
-> 1223     headers = build_hf_headers(
   1224         token=token,
   1225         library_name=library_name,

/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_validators.py in _inner_fn(*args, **kwargs)
    116             kwargs = smoothly_deprecate_use_auth_token(fn_name=fn.__name__, has_token=has_token, kwargs=kwargs)
    117 
--> 118         return fn(*args, **kwargs)
    119 
    120     return _inner_fn  # type: ignore

/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_headers.py in build_hf_headers(token, is_write_action, library_name, library_version, user_agent)
    119     """
    120     # Get auth token to send
--> 121     token_to_send = get_token_to_send(token)
    122     _validate_token_to_send(token_to_send, is_write_action=is_write_action)
    123 

/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_headers.py in get_token_to_send(token)
    146 
    147     # Token is not provided: we get it from local cache
--> 148     cached_token = get_token()
    149 
    150     # Case token is explicitly required

/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_token.py in get_token()
     39         `str` or `None`: The token, `None` if it doesn't exist.
     40     """
---> 41     return _get_token_from_google_colab() or _get_token_from_environment() or _get_token_from_file()
     42 
     43 

/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_token.py in _get_token_from_google_colab()
     57 
     58     try:
---> 59         token = userdata.get("HF_TOKEN")
     60     except userdata.NotebookAccessError:
     61         # Means the user has a secret call `HF_TOKEN` and got a popup "please grand access to HF_TOKEN" and refused it

/usr/local/lib/python3.10/dist-packages/google/colab/userdata.py in get(key)
     33     SecretNotFoundError: If the requested secret is not found.
     34   """
---> 35   resp = _message.blocking_request(
     36       'GetSecret', request={'key': key}, timeout_sec=None
     37   )

/usr/local/lib/python3.10/dist-packages/google/colab/_message.py in blocking_request(request_type, request, timeout_sec, parent)
    174       request_type, request, parent=parent, expect_reply=True
    175   )
--> 176   return read_reply_from_input(request_id, timeout_sec)

/usr/local/lib/python3.10/dist-packages/google/colab/_message.py in read_reply_from_input(message_id, timeout_sec)
     94     reply = _read_next_input_message()
     95     if reply == _NOT_READY or not isinstance(reply, dict):
---> 96       time.sleep(0.025)
     97       continue
     98     if (

KeyboardInterrupt:

System info

- huggingface_hub version: 0.20.1
- Platform: Linux-6.2.0-39-generic-x86_64-with-glibc2.35
- Python version: 3.10.12
- Running in iPython ?: No
- Running in notebook ?: No
- Running in Google Colab ?: No
- Token path ?: /home/tanaga/.cache/huggingface/token
- Has saved token ?: True
- Who am I ?: tanaga9
- Configured git credential helpers: store
- FastAI: 2.7.13
- Tensorflow: 2.10.1
- Torch: 2.1.0+cu118
- Jinja2: 3.1.2
- Graphviz: 0.20.1
- Pydot: 1.4.2
- Pillow: 9.4.0
- hf_transfer: N/A
- gradio: 4.12.0
- tensorboard: N/A
- numpy: 1.23.5
- pydantic: 2.5.3
- aiohttp: 3.8.6
- ENDPOINT: https://huggingface.co
- HF_HUB_CACHE: /home/tanaga/.cache/huggingface/hub
- HF_ASSETS_CACHE: /home/tanaga/.cache/huggingface/assets
- HF_TOKEN_PATH: /home/tanaga/.cache/huggingface/token
- HF_HUB_OFFLINE: False
- HF_HUB_DISABLE_TELEMETRY: False
- HF_HUB_DISABLE_PROGRESS_BARS: None
- HF_HUB_DISABLE_SYMLINKS_WARNING: False
- HF_HUB_DISABLE_EXPERIMENTAL_WARNING: False
- HF_HUB_DISABLE_IMPLICIT_TOKEN: False
- HF_HUB_ENABLE_HF_TRANSFER: False
- HF_HUB_ETAG_TIMEOUT: 10
- HF_HUB_DOWNLOAD_TIMEOUT: 10

[Wauplin]

Hi @tanaga9, sorry you're facing this issue. Thanks for reporting it. Could you tell me more about the environment you're using? I'm asking since if you reach line ---> 59 token = userdata.get("HF_TOKEN"), it means that:

1. huggingface_hub detects that you are running in a google colab here
   • => meaning that str(get_ipython()) contains "google.colab" as a string (see here)
2. that google.colab.userdata is installed locally (see here)
3. but that google.colab.userdata.get("HF_TOKEN") doesn't work.
   • => which looks more like an issue with google.colab package itself

What's also weird is that when you run huggingface-cli env it looks like it doesn't detect you as running in a colab (see in the dump you copy-pasted: - Running in Google Colab ?: No). I'm in favor of having a fix in huggingface_hub to prevent this if that's possible but we would need to understand better what's going on here. Thanks in advance.

[tanaga9]

hi @Wauplin , Thank you for your reply.
I am using the Google Colab image on Docker on Ubuntu.
image name is asia-docker.pkg.dev/colab-images/public/runtime.
https://console.cloud.google.com/artifacts/docker/colab-images/asia/public/runtime

By setting the FROM to that image and the CMD to jupyter lab ..., I can use the same instance for both local JupyterLab and as a local runtime from Colab.

I tried to create separate images, but it was too much work, so I unified them.

In this state, get_ipython() returned google.colab._shell.Shell, even when using it as a local JupyterLab (without using the Colab interface).

[Wauplin]

Hi @tanaga9, thanks for the context. I wasn't aware of this runtime.

I have been able to reproduce the issue and created PR #1953 to fix it. The plan is to make a hot-fix release once it's merged and make it the default in google colab. In the meantime, a temporary solution is to install a different version of huggingface_hub before executing any code. More details in #1953 (comment).

[tanaga9]

thanks

There is no problem as there is already a workaround.
As a precaution, I have attached the link to the document I saw.

• Local runtimes - Google Colab

[Wauplin]

#1953 is merged and the fix has been released (0.20.2). I'm closing this issue.

[Wauplin]

Update: huggingface_hub 0.20.2 is now the default version on Google Colab, meaning this issue should be fixed for everyone without having to upgrade (see googlecolab/colabtools#4272 (comment)).