OpenGVLab / LAMM

[NeurIPS 2023 Datasets and Benchmarks Track] LAMM: Multi-Modal Large Language Models and Applications as AI Agents
https://openlamm.github.io/
286 stars 15 forks source link

how to deal with multi-turn dialogue for octivius? #72

Open joez17 opened 5 months ago

joez17 commented 5 months ago

It seems that in Octivius, lora-moe uses conversation[0]['value'] to obtain the soft_gate value. image There are 2 questions: 1 Where are the system message and modality embedding introduced into gate activation? image

  1. In the case of multi-turn dialogues, incorporating only the initial question for gate computation throughout the entire conversation seems illogical.

Could there be aspects I'm misunderstanding? Please help clarify my confusion. Thanks!