hyper-p.json

{
    "data": {
        "dev": "dev_fr",
        "test": "test_fr",
        "filter": "10:1500"
    },
    "tokenizer": {
        "type": "LexiconTokenizer",
        "option-init": {
            "lexicon": "dict/fr/lexicon.txt"
        },
        "file": "dict/fr/tokenizer.tknz",
        "|V|": 35
    },
    "train": {
        "bin": "cat.ctc.train",
        "option": {
            "amp": true,
            "batch_size": 480,
            "check_freq": 500,
            "verbose": true,
            "large-dataset": true,
            "ld": "data/fr/{11_1000,1001_1200}/*.tar"
        }
    },
    "inference": {
        "avgmodel": {
            "mode": "best",
            "num": 3
        },
        "infer": {
            "bin": "cat.ctc.decode",
            "option": {
                "beam_size": 16,
                "nj": 16,
                "store_ark": true
            }
        },
        "er": {
            "per": true
        }
    },
    "commit": "1b77bc21f7a5faac5e47f6bc4445f23ebf0e0633"
}