Quantization Support for Gliner Model during Fine-Tuning on Custom Data

[Agrim1210]

I am fine-tuning the [GLiNER-medium-v2.1] on custom data and need to apply quantization for performance optimization. I am looking for guidance on:

Supported Quantization Methods:

1. Which methods (e.g., Post-training, Quantization-Aware Training) are compatible during fine-tuning?
2. Workflow: How can I integrate quantization into the fine-tuning pipeline? Are there recommended tools or best practices?
3. Limitations: Any known issues or limitations specific to quantizing the Gliner model?
4. Example: Can you provide a sample implementation or reference for applying quantization in this context?