* add pixtral text model (vision is wip) * cgraph ok, just missing 2D RoPE * fix bad rebase * first working version * fix problem with img_break token * support dynamic image size * update docs * update test script
* llava : update documentations * fix typo