import os import shutil import glob import random DIR = "/gpfs/u/home/LMCG/LMCGljnn/scratch-shared/junyan/raw" OUT_DIR = "/gpfs/u/home/LMCG/LMCGljnn/scratch-shared/junyan/raw/blip2_mini_dataset_full_karpathy" if __name__ == "__main__": os.makedirs(OUT_DIR, exist_ok=True) cc3m_tars = glob.glob(os.path.join(DIR, "cc3m", "cc3m_*", "*.tar")) cc12m_tars = glob.glob(os.path.join(DIR, "cc12m", "tars", "*.tar")) coco_tars = glob.glob(os.path.join(DIR, "karpathy_coco_wds_full", "*.tar")) vg_tars = glob.glob(os.path.join(DIR, "vg_wds_full", "*.tar")) tars = [] tars.extend(cc3m_tars) tars.extend(cc12m_tars) tars.extend(coco_tars) tars.extend(vg_tars) random.shuffle(tars) for i, tar in enumerate(tars): dst = os.path.join(OUT_DIR, f"{str(i).zfill(6)}.tar") print(tar, dst) os.symlink(tar, dst)