ocr/ocr_config.json

{
    "alphabet": " !\"'()*+,-./0123456789:;<=>?ABCDEFGHIJKLMNOPRSTVWY[\\]_abcdefghiklmnoprstuvwxyz|}ЁАБВГДЕЖЗИКЛМНОПРСТУФХЦЧШЩЫЬЭЮЯабвгдежзийклмнопрстуфхцчшщъыьэюяё’№",
    "save_dir": "data/experiments/tetradi_64_512_exp2/",
    "num_epochs": 300,
    "pretrain_path": "data/pretrain/model-187-0.6814.ckpt",
    "image": {
        "width": 512,
        "height": 64
    },
    "train": {
        "datasets": [
            {
                "csv_path": "data/school_notebooks_EN/train/train.csv",
                "prob": 14595
            },
            {
                "csv_path": "data/school_notebooks_RU/train/train.csv",
                "prob": 268350
            }
        ],
        "epoch_size": 100000,
        "batch_size": 512
    },
    "val": {
        "datasets": [
            {
                "csv_path": "data/school_notebooks_EN/val/val.csv",
                "prob": 2129
            },
            {
                "csv_path": "data/school_notebooks_RU/val/val.csv",
                "prob": 27853
            }
        ],
        "epoch_size": null,
        "batch_size": 512
    },
    "test": {
        "datasets": [
            {
                "csv_path": "data/school_notebooks_EN/test/test.csv",
                "prob": 1979
            },
            {
                "csv_path": "data/school_notebooks_RU/test/test.csv",
                "prob": 28109
            }
        ],
        "epoch_size": null,
        "batch_size": 512
    }
}