example_data_config.yaml

# Data args.
# Note: this is an example data config, not for reproducing xgen-mm-instruct.
data_path: {

  #'/blip-3/dataset/blip_laion_cc_sbu_558k_fixed.json': 558128
  #'/blip-3/dataset/LLaVA-Pretrain/blip_laion_cc_sbu_558k_fixed.json': 558128
  # '/mnt/xgen-mm/LLaVA-Pretrain/llava_all_path.json': 558128 

  # # Llava-665K
  # '/export/home/blip3_data/llava_instruct_665k_sharegpt4v/annotations/sharegpt4v_mix665k_cap23k_coco-ap9k_lcs3k_sam9k_div2k.json': 665058, # Total: 665058,
  # SoM-llava.
  '/blip-3_pytorch/dataset/SoM-LLaVA/som_qa_coco20k.json': 20160,
  '/blip-3_pytorch/dataset/SoM-LLaVA/som_listing_coco10k.json': 10000,

  # # Text-only. (37k)
  # # '/export/share/manlis/data/allava-4v/Evol-Instruct-GPT4-Turbo-143K-filterd.json': 20000,  # Total: 143000
  # '/export/home/blip3_data/text-only-sft-data/Python-Code-23k-ShareGPT.json': 10000, # Total 22608
  # '/export/home/blip3_data/text-only-sft-data/gsm8k-main-train.json': 7473,
  # '/export/home/blip3_data/text-only-sft-data/slimorca-dedup.json': 10000, # Total: 363491
  # '/export/home/blip3_data/text-only-sft-data/orca-math-word-problems-200k.json': 10000, # Total: 200035
  # '/export/home/blip3_data/text-only-sft-data/lima-train.json': 5000, #Total: 1030

  # # OCR (72k)
  # '/export/home/blip3_data/ocr_datasets/ai2d/ai2d_multichoice_llava_format_single_img_token_train.json': 10000, # Total: 2482
  # '/export/home/blip3_data/ocr_datasets/DVQA/dvqa_llava_format.json': 20000, # Total: 2325316
  # '/export/home/blip3_data/ocr_datasets/DocVQA/docvqa_llava_format.json': 20649,
  # '/export/home/blip3_data/ocr_datasets/chartQA/chartqa_train_augmented_llava_format.json': 20901,
  # '/export/home/blip3_data/ocr_datasets/chartQA/chartqa_train_human_llava_format.json': 7398,

}