[Question] Does all LLaVA1.5 dataset opensourced?

[OpenJarvisAI]

Question

In paper, it says:

1. subset of some public dataset, is there any plan to open the sampled dataset for reproduce?
2. you said the training are hard to converge with reduandt params and limited data, why not add all data?