Installation¶

To begin using the GLiNER model, you can install the GLiNER Python library through pip, conda, or directly from the source.

Install via Pip¶

pip install gliner

ONNX Runtime is optional. Choose the onnx extra for CPU inference or the gpu extra for CUDA execution providers. Install only one: the CPU and GPU packages share the same Python module. Install the separate onnx package to export models.

pip install "gliner[onnx]"  # CPU-backed ONNX Runtime
# or
pip install "gliner[gpu]"   # GPU-backed ONNX Runtime

pip install onnx           # additionally required for ONNX export

To export or run models with OpenVINO on CPU, GPU, or NPU devices supported by OpenVINO, install the optional OpenVINO dependency.

pip install "gliner[openvino]"

See ONNX Runtime and OpenVINO for model conversion and runtime configuration.

Install via Conda¶

conda install -c conda-forge gliner

Install from Source¶

To install the GLiNER library from source, follow these steps:

  1. Clone the Repository:

    First, clone the GLiNER repository from GitHub:

    git clone https://github.com/urchade/GLiNER

  2. Navigate to the Project Directory:

    Change to the directory containing the cloned repository:

    cd GLiNER

  3. Install Dependencies:

    It’s a good practice to create and activate a virtual environment before installing dependencies:

    python -m venv venv
source venv/bin/activate   # On Windows use: venv\Scripts\activate

    Install the required dependencies listed in the requirements.txt file:

    pip install -r requirements.txt

  4. Install the GLiNER Package:

    Finally, install the GLiNER package using the setup script:

    pip install .

  5. Verify Installation:

    You can verify the installation by importing the library in a Python script:

    import gliner
print(gliner.__version__)

Install FlashDeBERTa¶

Most GLiNER models use the DeBERTa encoder as their backbone. This architecture offers strong token classification performance and typically requires less data to achieve good results. However, a major drawback has been its slower inference speed, and until recently, there was no flash attention implementation compatible with DeBERTa’s disentangled attention mechanism.

To address this, FlashDeBERTa was introduced.

To use FlashDeBERTa with GLiNER, install it:

pip install flashdeberta -U

Before using FlashDeBERTa, please make sure that you have transformers>=4.51.3.

To enable FlashDeBERTa, set the USE_FLASHDEBERTA environment variable before loading the model:

export USE_FLASHDEBERTA=1

Or set it directly in Python:

import os
os.environ["USE_FLASHDEBERTA"] = "1"

from gliner import GLiNER
model = GLiNER.from_pretrained("urchade/gliner_mediumv2.1")

If needed, you can switch to the standard eager attention mechanism by specifying the attention implementation:

model = GLiNER.from_pretrained("urchade/gliner_mediumv2.1", _attn_implementation="eager")

FlashDeBERTa provides up to a 3× speed boost for typical sequence lengths—and even greater improvements for longer sequences. alt text