Train dataset

[egbertYeah]

thanks for your great work. i have some question about training dataset.

1. Are these datasets (as shown in the summary image) simply concatenated into a single file (viscot_mixed_2m.json) and then directly used for the second stage of training ?
2. The Visual CoT dataset contains 438k samples，which utilized in any time?

[egbertYeah]

"gqa_cot_train.jsonl" this file is used to training model?