Jñāna is a multimodal LLM gradio app hosted on huggingface spaces. It is capable of accepting inputs in the form of image/audio/text or a combination of any of these 3. Jñāna uses microsoft/phi2 LLM model that was trained based on Llava 1.0 and Llava 1.5. qlora strategy was used for fine-tuning microsoft/phi2.