{
    "_name_": "SequenceClassifier",
    "_model_version_": "1.0.0",
    "_symupe_version_": "1.1.0",
    "backbone": {
        "dim": 768,
        "token_embeddings": {
            "_target_": "simple",
            "emb_dims": 128,
            "mode": "cat",
            "emb_norm": true,
            "discrete": false,
            "sinusoidal": true,
            "sinusoidal_learned": true,
            "project_bias": false,
            "embedding_kwargs": {
                "depth": 0,
                "log_inv_freq": true,
                "with_positions": false
            },
            "tie_keys": null,
            "special_tokens": {
                "PAD_None": 0,
                "MASK_None": 1,
                "BOS_None": 2,
                "EOS_None": 3,
                "IGNORE_None": 4,
                "EOD_None": 5,
                "Bar_Line": 6,
                "Pedal_On": 7,
                "Pedal_Off": 8
            }
        },
        "emb_norm": false,
        "emb_dropout": 0.0,
        "use_abs_pos_emb": false,
        "project_bias": false,
        "transformer": {
            "_target_": "encoder",
            "depth": 12,
            "memory_tokens": 4,
            "final_norm_bias": false,
            "attention": {
                "_target_": "global",
                "heads": 12,
                "head_dim": 64,
                "one_kv_head": true,
                "dropout": 0.05,
                "rotary_pos_emb": true,
                "rotary_emb_base": 1024,
                "alibi_pos_bias": false,
                "alibi_learned": true,
                "alibi_heads": 8
            },
            "feed_forward": {
                "mult": 3,
                "glu": true,
                "swish": true,
                "dropout": 0.05
            }
        },
        "lm_head": null,
        "num_tokens": {
            "Pitch": 97,
            "TimeShift": 595,
            "Velocity": 42,
            "TimeDurationSustain": 560,
            "TimePosition": 2409
        }
    },
    "backbone_checkpoint": null,
    "transformer": {
        "dim": 128,
        "depth": 1,
        "final_norm_bias": false,
        "attention": {
            "_target_": "global",
            "heads": 2,
            "head_dim": 64,
            "one_kv_head": true,
            "dropout": 0.2
        },
        "feed_forward": {
            "mult": 1,
            "glu": false,
            "swish": true,
            "dropout": 0.2
        }
    },
    "classifier": {
        "hidden_dims": null
    },
    "aggregation": "token",
    "emb_norm": false,
    "emb_dropout": 0.3,
    "clf_dropout": 0.3,
    "context_with_memory": true,
    "backbone_output_layer": 10,
    "num_classes": 4,
    "note_classifier": false,
    "label_smoothing": 0.0,
    "detach_inputs": true,
    "memory_norm": false,
    "labels": {
        "0": "score",
        "1": "high quality",
        "2": "low quality",
        "3": "corrupted"
    }
}