Installation

GLiFormer requires Python 3.10 or newer. Install the framework from PyPI:

pip install gliformer

The installation includes PyTorch, Transformers, and Pydantic. CPU inference uses eager attention. To install the optional CUDA attention kernels:

pip install "gliformer[flash]"

Optional Dependencies

Install extras for the features you need:

Extra	Purpose
flash	FlashDeBERTa attention support
data	Dataset loading with Hugging Face Datasets
pdf	PDF processing with PyMuPDF and pdfplumber
demo	Gradio demos and comparison models
vision	Torchvision utilities
audio	Torchaudio utilities
dev	pytest and Ruff
all	All optional dependencies

For example, to install the demo and PDF utilities:

pip install "gliformer[demo,pdf]"

Optional packages supply framework utilities; tasks also require a checkpoint trained with the corresponding heads.

Continue with the Quickstart to load Base v1 or Large v1.