A development environment for Transformer models, embeddings, and LLM integrations. This project leverages uv for high-performance dependency management.
Ensure you have uv installed:
curl -LsSf https://astral-sh.uv/install.sh | shSync the environment and install all dependencies:
uv sync- Transformers & Tokenizers: Hugging Face libraries for NLP.
- LangChain: Integrations for OpenAI and Google GenAI.
- MTEB: Benchmarking for text embeddings.
- Sentence-Transformers: Simplified embedding generation.
- PyTorch: Deep learning backend.
A serious vulnerability exists in torch.load for PyTorch versions prior to v2.6.0. This vulnerability allows for arbitrary code execution when loading untrusted .pt or .bin files.
Action Required:
The project dependencies have been updated to require torch >= 2.6.0.
If you must work with older checkpoints, ensure you use the safetensors format, which is immune to this vulnerability:
from transformers import AutoModel
model = AutoModel.from_pretrained("model_name", use_safetensors=True)