## Quantize steps

```bash
optimum-cli export onnx --model vikp/texify --task image-to-text-with-past onnx/
optimum-cli onnxruntime quantize --onnx_model onnx/ --avx2 -o quantized_model/
```

> Remember to use the `preprocess_config.json` from https://huggingface.co/Xenova/texify