{
    "train_datasets": [
        {"name": "coco_cap",
         "db": ["/db/pretrain_caption_coco_train_base-cased.db/",
                "/db/pretrain_caption_coco_trainval_base-cased.db/"],
         "img": ["/img/coco_train2014/", "/img/coco_val2014/"],
         "tasks": ["mrm-nce"],
         "mix_ratio": [1]}
    ],
    "val_datasets": [
        {"name": "coco_cap",
         "db": ["/db/pretrain_caption_coco_val_base-cased.db/"],
         "img": ["/img/coco_val2014/"],
         "tasks": ["mrm-nce"]}
    ],
    "output_dir": "/storage/pretrain/mrm_nce_coco",
    "mrm_prob": 0.15,
    "neg_size": 1024,
    "nce_temp": 1.0,
    "itm_neg_prob": 0.5,
    "max_txt_len": 60,
    "conf_th": 0.2,
    "max_bb": 100,
    "min_bb": 10,
    "num_bb": 36,
    "train_batch_size": 8192,
    "val_batch_size": 8192,
    "gradient_accumulation_steps": 2,
    "learning_rate": 5e-05,
    "valid_steps": 5000,
    "num_train_steps": 100000,
    "optim": "adamw",
    "decay": "linear",
    "dropout": 0.1,
    "weight_decay": 0.01,
    "grad_norm": 2.0,
    "warmup_steps": 10000,
    "seed": 42,
    "fp16": true,
    "pin_mem": true,
    "n_workers": 4,
    "from_scratch": false
}