3 ms·

You probably don't want to fine-tune a quantized model. They are fine for inference but not great for training.