---
license: apache-2.0
---
**Base Model**: BLIP2-t5 pretrained version
**Finetune data**: LLAVA 150k (sample one pair of instruction-answer if multi-round conversations)
**Hyper-parameters**:
* lr = 2e-5
* gbs = 32
* image size = 480
* weight decay = 0.05
| 
