{ "train_txt_dbs": ["/db/itm_flickr30k_train_base-cased.db"], "train_img_dbs": ["/img/flickr30k/"], "val_txt_db": "/db/itm_flickr30k_val_base-cased.db", "val_img_db": "/img/flickr30k/", "test_txt_db": "/db/itm_flickr30k_test_base-cased.db", "test_img_db": "/img/flickr30k/", "checkpoint": "/pretrain/alltask_ot_alldata.pt", "model_config": "/src/config/uniter-base.json", "output_dir": "/storage/finetune/itm/flickr_ot_alldata_base_hnv2", "max_txt_len": 60, "conf_th": 0.2, "max_bb": 100, "min_bb": 10, "num_bb": 36, "train_batch_size": 8, "negative_size": 399, "hard_neg_size": 31, "inf_minibatch_size": 400, "learning_rate": 5e-05, "valid_steps": 500, "num_train_steps": 5000, "optim": "adamw", "betas": [ 0.9, 0.98 ], "margin": 0.2, "dropout": 0.1, "weight_decay": 0.01, "grad_norm": 2.0, "warmup_steps": 500, "seed": 42, "fp16": true, "n_workers": 4, "pin_mem": true, "full_val": true }