Add Sentio (WeMM-Embedding-9B + LoRA, Qwen2.5-Omni-3B audio tower + LoRA)

#197
by Dragos-Datcu - opened

Sentio: WeMM-Embedding-9B with a LoRA fine-tune for video moment retrieval and an audio branch (Qwen2.5-Omni-3B audio tower with LoRA, trained connector/adapter head), 10.20B parameters, 4096-d. All 190 MMEB-V3 tasks evaluated with the Tencent WeMM harness (audio tasks with the Wianmu audio modifications). Model card: https://huggingface.co/Dragos-Datcu/Sentio

ziyjiang changed pull request status to merged

Sign up or log in to comment