How to evaluate your LLaVA-UHD model?

[Gaffey]

It is said that you use the same scripts as LLaVA in README. However, all codes about UHD are writen in llava/train/llava_uhd, but the original script cannot use these code since llava/model/builder.py did not even import your UHD part. Could you please provide the corresponding evaluation code? Directly import these codes cannot make it since the data preprocesser is not correct.

[guozonghao96]

Our repository has been fully improved, and almost all bugs have been eliminated. For details, please refer to the main branch and the LLaVA-UHD v1 branch.

You can use the scripts to evaluate the model like llava-uhd v1 or v2. The eval datasets can be downloaded like llava-1.5. Besides, you can also use the VLMEval-kit to evaluate all the benchmark.