search

huggingface / huggingface_hub

The official Python client for the Huggingface Hub.
https://huggingface.co/docs/huggingface_hub
Apache License 2.0
2.13k stars 557 forks source link

transformers dependency_versions_check.py causes circular import when trying to use transformers.LlamaTokenizerFast #2562

Open m9e opened 1 month ago

m9e commented 1 month ago

Describe the bug

Trying to load LlamaTokenizerFast in code like this:

import transformers

# Load the tokenizer (replace 'bert-base-uncased' with the tokenizer you prefer)
#tokenizer = transformers.LlamaTokenizerFast.from_pretrained('/var/models/meta-llama/Meta-Llama-3-8B-Instruct/')
tokenizer = transformers.LlamaTokenizerFast.from_pretrained('meta-llama/Meta-Llama-3-8B-Instruct')

# Predefine your favorite sayings
sayings = [
    "The quick brown fox",
    "jumps over the lazy dog",
    "Carpe diem",
    "Veni vidi vici"
]

# Tokenize each phrase into a list of tokens
tokenized_sayings = [tokenizer.encode(phrase, add_special_tokens=False) for phrase in sayings]

# Print out the tokenized representations
for idx, tokenized_phrase in enumerate(tokenized_sayings):
    print(f"Phrase {idx+1}: {tokenized_phrase}")

Results in this:

Error importing huggingface_hub.hf_api: partially initialized module 'transformers' has no attribute 'LlamaTokenizerFast' (most likely due to a circular import)
Error importing huggingface_hub.hf_api: partially initialized module 'transformers' has no attribute 'LlamaTokenizerFast' (most likely due to a circular import)
Traceback (most recent call last):
  File "/Users/matt/code/hftest/tokenize.py", line 1, in <module>
    import transformers
  File "/Users/matt/code/hftest/venv/lib/python3.10/site-packages/transformers/__init__.py", line 26, in <module>
    from . import dependency_versions_check
  File "/Users/matt/code/hftest/venv/lib/python3.10/site-packages/transformers/dependency_versions_check.py", line 16, in <module>
    from .utils.versions import require_version, require_version_core
  File "/Users/matt/code/hftest/venv/lib/python3.10/site-packages/transformers/utils/__init__.py", line 18, in <module>
    from huggingface_hub import get_full_repo_name  # for backward compatibility
ImportError: cannot import name 'get_full_repo_name' from 'huggingface_hub' (/Users/matt/code/hftest/venv/lib/python3.10/site-packages/huggingface_hub/__init__.py)

Reproduction

mikoshi:code matt$ mkdir hftest
mikoshi:code matt$ cd hftest
mikoshi:hftest matt$ python --version
Python 3.10.14
mikoshi:hftest matt$ python -m venv venv
mikoshi:hftest matt$ source venv/bin/activate
(venv) mikoshi:hftest matt$ pip install transformers huggingface_hub
Collecting transformers
  Using cached transformers-4.44.2-py3-none-any.whl (9.5 MB)
[snip]
(venv) mikoshi:hftest matt$ vi tokenize.py 
(venv) mikoshi:hftest matt$ !py
python tokenize.py 
Error importing huggingface_hub.hf_api: partially initialized module 'transformers' has no attribute 'LlamaTokenizerFast' (most likely due to a circular import)
Error importing huggingface_hub.hf_api: partially initialized module 'transformers' has no attribute 'LlamaTokenizerFast' (most likely due to a circular import)
Traceback (most recent call last):
  File "/Users/matt/code/hftest/tokenize.py", line 1, in <module>
[snip]

I've played with a bunch of downgrades here trying to do hf hub <0.25 or transformers <4.44 but I get the same result on all of them.

First batch was OSX, identical result on Ubuntu 22.04 LTS+python 3.10.12:

(venv) matt@firestorm:~/nvme/code/hftest$ python tokenize.py 
Error importing huggingface_hub.hf_api: partially initialized module 'transformers' has no attribute 'LlamaTokenizerFast' (most likely due to a circular import)
Error importing huggingface_hub.hf_api: partially initialized module 'transformers' has no attribute 'LlamaTokenizerFast' (most likely due to a circular import)
Traceback (most recent call last):
  File "/home/matt/nvme/code/hftest/tokenize.py", line 1, in <module>
    import transformers
  File "/home/matt/nvme/code/hftest/venv/lib/python3.10/site-packages/transformers/__init__.py", line 26, in <module>
    from . import dependency_versions_check
  File "/home/matt/nvme/code/hftest/venv/lib/python3.10/site-packages/transformers/dependency_versions_check.py", line 16, in <module>
    from .utils.versions import require_version, require_version_core
  File "/home/matt/nvme/code/hftest/venv/lib/python3.10/site-packages/transformers/utils/__init__.py", line 18, in <module>
    from huggingface_hub import get_full_repo_name  # for backward compatibility
ImportError: cannot import name 'get_full_repo_name' from 'huggingface_hub' (/home/matt/nvme/code/hftest/venv/lib/python3.10/site-packages/huggingface_hub/__init__.py)
(venv) matt@firestorm:~/nvme/code/hftest$ uname -a
Linux firestorm 6.8.0-40-generic #40~22.04.3-Ubuntu SMP PREEMPT_DYNAMIC Tue Jul 30 17:30:19 UTC 2 x86_64 x86_64 x86_64 GNU/Linux
(venv) matt@firestorm:~/nvme/code/hftest$

Logs

No response

System info

- huggingface_hub version: 0.25.1
- Platform: macOS-14.6.1-arm64-arm-64bit
- Python version: 3.10.14
- Running in iPython ?: No
- Running in notebook ?: No
- Running in Google Colab ?: No
- Running in Google Colab Enterprise ?: No
- Token path ?: /Users/matt/.cache/huggingface/token
- Has saved token ?: True
- Who am I ?: m9e
- Configured git credential helpers: osxkeychain
- FastAI: N/A
- Tensorflow: N/A
- Torch: N/A
- Jinja2: N/A
- Graphviz: N/A
- keras: N/A
- Pydot: N/A
- Pillow: N/A
- hf_transfer: N/A
- gradio: N/A
- tensorboard: N/A
- numpy: 2.1.1
- pydantic: N/A
- aiohttp: N/A
- ENDPOINT: https://huggingface.co
- HF_HUB_CACHE: /Users/matt/.cache/huggingface/hub
- HF_ASSETS_CACHE: /Users/matt/.cache/huggingface/assets
- HF_TOKEN_PATH: /Users/matt/.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

linux version:

- huggingface_hub version: 0.25.1
- Platform: Linux-6.8.0-40-generic-x86_64-with-glibc2.35
- Python version: 3.10.12
- Running in iPython ?: No
- Running in notebook ?: No
- Running in Google Colab ?: No
- Running in Google Colab Enterprise ?: No
- Token path ?: /home/matt/.cache/huggingface/token
- Has saved token ?: True
- Who am I ?: m9e
- Configured git credential helpers: cache
- FastAI: N/A
- Tensorflow: N/A
- Torch: N/A
- Jinja2: N/A
- Graphviz: N/A
- keras: N/A
- Pydot: N/A
- Pillow: N/A
- hf_transfer: N/A
- gradio: N/A
- tensorboard: N/A
- numpy: 2.1.1
- pydantic: N/A
- aiohttp: N/A
- ENDPOINT: https://huggingface.co
- HF_HUB_CACHE: /home/matt/.cache/huggingface/hub
- HF_ASSETS_CACHE: /home/matt/.cache/huggingface/assets
- HF_TOKEN_PATH: /home/matt/.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
m9e commented 1 month ago

I should add, I submitted here because I think the root cause is, as you can see in the errors:

ImportError: cannot import name 'get_full_repo_name' from 'huggingface_hub' (/home/matt/nvme/code/hftest/venv/lib/python3.10/site-packages/huggingface_hub/__init__.py)

hanouticelina commented 1 month ago

Hi @m9e 👋 I've tried to reproduce the issue using the same versions of transformers==4.44.2 and huggingface_hub==0.25.1, but I was not able to replicate the error. I suspect it may be related to package compatibility or maybe installation issues rather than a bug in the code itself. You can try creating a new virtual env and then upgrade to the latest versions pip install -U transformers huggingface_hub