config.json

{
  "architectures": [
    "CM3PForBeatmapClassification"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "audio_config": {
    "attention_bias": false,
    "attention_dropout": 0.0,
    "decoder_bias": true,
    "deterministic_flash_attn": false,
    "embedding_dropout": 0.0,
    "f_max": 8000,
    "f_min": 0,
    "global_attn_every_n_layers": 3,
    "global_rope_theta": 160000.0,
    "hidden_activation": "gelu",
    "hidden_size": 512,
    "hop_length": 128,
    "initializer_cutoff_factor": 2.0,
    "initializer_range": 0.02,
    "intermediate_size": 1024,
    "local_attention": 128,
    "local_rope_theta": 10000.0,
    "max_position_embeddings": 4096,
    "mlp_bias": false,
    "mlp_dropout": 0.0,
    "model_type": "CM3PAudio",
    "n_ftt": 2048,
    "n_mels": 80,
    "norm_bias": false,
    "norm_eps": 1e-05,
    "num_attention_heads": 8,
    "num_hidden_layers": 6,
    "pad_mode": "constant",
    "projector_dim": 768,
    "projector_hidden_act": "gelu",
    "projector_intermediate_size": 2048,
    "sample_rate": 16000,
    "torch_dtype": "bfloat16",
    "vocab_size": 1
  },
  "audio_eos_token_id": 3966,
  "audio_sos_token_id": null,
  "audio_token_id": 3967,
  "auto_map": {
    "AutoConfig": "configuration_cm3p.CM3PBeatmapConfig",
    "AutoModelForSequenceClassification": "modeling_cm3p.CM3PForBeatmapClassification"
  },
  "bos_token_id": 3958,
  "classifier_activation": "gelu",
  "classifier_bias": false,
  "cls_embed": true,
  "decoder_bias": true,
  "deterministic_flash_attn": false,
  "embedding_dropout": 0.0,
  "eos_token_id": 3959,
  "global_attn_every_n_layers": 3,
  "global_rope_theta": 160000.0,
  "hidden_activation": "gelu",
  "hidden_size": 768,
  "id2label": {
    "0": "Graveyard",
    "1": "Ranked"
  },
  "initializer_cutoff_factor": 2.0,
  "initializer_factor": 1.0,
  "initializer_range": 0.02,
  "intermediate_size": 1152,
  "label2id": null,
  "local_attention": 128,
  "local_rope_theta": 10000.0,
  "max_position_embeddings": 8192,
  "mlp_bias": false,
  "mlp_dropout": 0.0,
  "model_type": "CM3PBeatmap",
  "norm_bias": false,
  "norm_eps": 1e-05,
  "num_attention_heads": 12,
  "num_hidden_layers": 22,
  "pad_token_id": 3962,
  "problem_type": "single_label_classification",
  "projection_dim": 512,
  "repad_logits_with_grad": false,
  "sparse_pred_ignore_index": -100,
  "sparse_prediction": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.55.0",
  "vocab_size": 3968
}