2FastVLM: Efficient vision encoding for vision language models (opens in new tab)(github.com)367nhod0y ago74