This is not a direct improvement of the v1 model, but rather a different variant, with a different dataset and training settings.
This is not a direct improvement of the v1 model, but rather a different variant, with a different dataset and training settings.