About the llava-ov single-image SFT data noted in single_image.yaml

[SAIS-FUXI]

I noted that the following data are noted in scripts/train/single_image.yaml, but not specified the reason why they are noted.
Are they used in the single-image SFT training of llava-ov models?

- json_path: /mnt/bn/vl-research/data/llava_instruct/real_vision_flan/sroie_data_33626.json

sampling_strategy: "first:10%"

- json_path: /mnt/bn/vl-research/data/llava_instruct/real_vision_flan/orand_car_a_train_2009.json

sampling_strategy: "all"

- json_path: /mnt/bn/vl-research/data/llava_instruct/real_vision_flan/orand_car_b_train_3000.json

sampling_strategy: "all"