config.json

{
    "version": "0.1",
    "model": {
        "encoder": {
            "n_layer": 12,
            "n_embd": 1024,
            "n_hidden": 4096,
            "n_head": 16,
            "head_dim": 128
        },
        "decoder": {
            "n_layer": 18,
            "n_embd": 2048,
            "n_hidden": 8192,
            "gqa_query_heads": 16,
            "cross_query_heads": 16,
            "kv_heads": 4,
            "gqa_head_dim": 128,
            "cross_head_dim": 128,
            "d_model" : 256
        },
        "src_vocab_size": 256,
        "tgt_vocab_size": 1028,
        "dropout": 0.0
    },
    "training": {
        "dtype": "bfloat16"
    },
    "data": {
        "text_length": 1024,
        "audio_length": 4096,
        "channels": 9,
        "text_pad_value": 0,
        "audio_eos_value": 1024,
        "audio_pad_value": 1025,
        "audio_bos_value": 1026,
        "delay_pattern": [
            0,
            8,
            9,
            10,
            11,
            12,
            13,
            14,
            15
        ]
    }
}