FireRedAudio/config.json

{
  "architectures": [
    "FireRedAudioForCausalLM"
  ],
  "audio_encoder_config": {
    "_attn_implementation_autoset": true,
    "activation_dropout": 0.1,
    "activation_function": "gelu",
    "add_cross_attention": false,
    "architectures": [
      "FireRed_AudioEncoder"
    ],
    "attention_dropout": 0.1,
    "bos_token_id": null,
    "cross_attention_hidden_size": null,
    "d_model": 1280,
    "decoder_start_token_id": null,
    "dropout": 0.1,
    "dtype": "bfloat16",
    "encoder_attention_heads": 20,
    "encoder_ffn_dim": 5120,
    "encoder_layers": 32,
    "eos_token_id": null,
    "finetuning_task": null,
    "init_std": 0.02,
    "initializer_range": 0.02,
    "is_decoder": false,
    "max_source_positions": 1500,
    "model_type": "firered_audio_encoder",
    "n_window": 1500,
    "num_mel_bins": 128,
    "output_dim": 4096,
    "pad_token_id": null,
    "prefix": null,
    "pruned_heads": {},
    "scale_embedding": false,
    "sep_token_id": null,
    "task_specific_params": null,
    "tf_legacy_loss": false,
    "tie_encoder_decoder": false,
    "tie_word_embeddings": true,
    "tokenizer_class": null,
    "torchscript": false,
    "use_bfloat16": false
  },
  "audio_special_no_latent_id": 248092,
  "audio_special_token": "<|AUDIO|>",
  "audio_special_token_id": 248091,
  "audio_special_token_no_latent": "<|AUDIO_NO_LATENT|>",
  "backbone_config": {
    "attention_bias": false,
    "attention_dropout": 0.1,
    "attn_output_gate": true,
    "bos_token_id": null,
    "dtype": "bfloat16",
    "eos_token_id": 248044,
    "full_attention_interval": 4,
    "head_dim": 256,
    "hidden_act": "silu",
    "hidden_size": 4096,
    "initializer_range": 0.02,
    "intermediate_size": 12288,
    "layer_types": [
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention"
    ],
    "linear_conv_kernel_dim": 4,
    "linear_key_head_dim": 128,
    "linear_num_key_heads": 16,
    "linear_num_value_heads": 32,
    "linear_value_head_dim": 128,
    "mamba_ssm_dtype": "float32",
    "max_position_embeddings": 262144,
    "mlp_only_layers": [],
    "model_type": "qwen3_5_text",
    "mtp_num_hidden_layers": 1,
    "mtp_use_dedicated_embeddings": false,
    "num_attention_heads": 16,
    "num_hidden_layers": 32,
    "num_key_value_heads": 4,
    "pad_token_id": null,
    "partial_rotary_factor": 0.25,
    "rms_norm_eps": 1e-06,
    "rope_parameters": {
      "mrope_interleaved": true,
      "mrope_section": [
        11,
        11,
        10
      ],
      "partial_rotary_factor": 0.25,
      "rope_theta": 10000000,
      "rope_type": "default"
    },
    "tie_word_embeddings": false,
    "use_cache": true,
    "vocab_size": 248320
  },
  "dit_config": {
    "backbone_hidden_size": 4096,
    "depth": 11,
    "hidden_size": 1024,
    "history_patches": 2,
    "mlp_ratio": 4.0,
    "model_type": "red_dit",
    "num_heads": 16,
    "patch_size": 4,
    "train_cfg_rate": 0.1,
    "vae_channels": 64
  },
  "dtype": "bfloat16",
  "eosp_idx": 248078,
  "model_type": "firered_audio",
  "patch_encoder_config": {
    "depth": 8,
    "hidden_size": 1024,
    "mlp_ratio": 4,
    "model_type": "red_patch_encoder",
    "num_heads": 16,
    "out_dim": 4096,
    "patch_size": 4,
    "semantic_dim": 0,
    "vae_dim": 64
  },
  "red_vae_config": {
    "audio_patch_size": 480,
    "audio_sample_rate": 24000,
    "downsample_num_hidden_layers": 4,
    "extra_downsample_rate": 2,
    "hidden_size": 896,
    "initializer_range": 0.02,
    "intermediate_size": 3584,
    "max_position_embeddings": 32768,
    "max_window_layers": 0,
    "model_type": "red_vae_audio_encoder_v1",
    "num_attention_heads": 14,
    "num_hidden_layers": 18,
    "num_key_value_heads": 2,
    "out_dim": 64,
    "sliding_window": 64,
    "use_sliding_window": true
  },
  "sosp_idx": 248077,
  "transformers_version": "5.8.0"
}