HuggingFaceEmbedding Unable to load local embedding model; Error importing huggingface_hub.hf_api: 'Version'

[tdecosta]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

from langchain_huggingface import HuggingFaceEmbeddings

device = 'cuda'#'cpu'

print(f"Initializing the Embedding Model: {embed_model_path}")
hfem = HuggingFaceEmbeddings(
    model_name="./models/sentence-transformers/all-MiniLM-L6-v2",
    model_kwargs={'device': device},
    encode_kwargs={'device': device, 'batch_size': 32}
)

Description

I recently updated a bunch pip packages to include langchain v3. After upgrading, I re-ran code (which worked previously) and resulted in an error when initializing a downloaded embedding model (sentence-transformers/all-MiniLM-L6-v2). Now I'm receiving a exception referencing a Key Error: 'Version'

I suspect there's some compatibility issue with huggingface-hub version? However, I'm not sure which version I should downgrade to be compatible with langchain-huggingface

System Info

System Information

OS: Windows
OS Version: 10.0.19045
Python Version: 3.12.4 (tags/v3.12.4:8e8a4ba, Jun 6 2024, 19:30:16) [MSC v.1940 64 bit (AMD64)]

Package Information

langchain_core: 0.3.13
langchain: 0.3.5
langchain_community: 0.3.3
langsmith: 0.1.137
langchain_chroma: 0.1.4
langchain_experimental: 0.3.2
langchain_huggingface: 0.1.1
langchain_text_splitters: 0.3.1
langchainhub: 0.1.21

Optional packages not installed

langgraph
langserve

Other Dependencies

aiohttp: 3.10.10
async-timeout: Installed. No version info available.
chromadb: 0.5.16
dataclasses-json: 0.6.7
fastapi: 0.111.0
httpx: 0.27.0
huggingface-hub: 0.26.2
jsonpatch: 1.33
numpy: 1.26.4
orjson: 3.10.10
packaging: 24.1
pydantic: 2.8.0
pydantic-settings: 2.6.0
PyYAML: 6.0.2
requests: 2.32.3
requests-toolbelt: 1.0.0
sentence-transformers: 3.2.1
SQLAlchemy: 2.0.36
tenacity: 8.5.0
tokenizers: 0.20.1
transformers: 4.46.1
types-requests: 2.32.0.20241016
typing-extensions: 4.12.2

[tdecosta]

After reviewing the call stack and diving down into the code of importlib, it became apparent there was an issue with obtaining the version installed for PyTorch. Turns out that if you have some lingering dist-info from previous installation of torch the importlib gets "confused" and return None for the version. That along with noticing that I had torch installed for the user and globally that conflicted with each other. After uninstalling torch and then manually doing some clean up on the global site-packages, then re-installing torch seemed to fix the issue.

This was not a compatibility issue between langchain-huggingface and huggingface-hub