Open chancharikmitra opened 11 months ago
We will release the evaluation code soon.
The evaluation code is available in https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/accessory/eval_mm, you may refer to the Readme.md for detailed instructions.
This seems like a great contribution to the MLLM space! I see that your model was evaluated on SEEDBench v1. Would you be able to share the exact scripts and prompts used for evaluation to replicate the same results? It would be great to have this script for use on other QA datasets as well.
Also, is Sphinx-v2-1k using the LLaMA2-13B LLM backbone?