{
  "ats_weight" : 0.5,
  "dtype" : "float16",
  "fc_encoder_config" : {
    "attention_bias" : 1,
    "conv_kernel_size" : 9,
    "hidden_act" : "silu",
    "hidden_size" : 512,
    "intermediate_size" : 2048,
    "max_position_embeddings" : 5000,
    "model_type" : "sortformer_fc_encoder",
    "num_attention_heads" : 8,
    "num_hidden_layers" : 17,
    "num_key_value_heads" : 8,
    "num_mel_bins" : 128,
    "scale_input" : 1,
    "subsampling_conv_channels" : 256,
    "subsampling_conv_kernel_size" : 3,
    "subsampling_conv_stride" : 2,
    "subsampling_factor" : 8
  },
  "model_type" : "sortformer",
  "modules_config" : {
    "causal_attn_rc" : 7,
    "chunk_left_context" : 1,
    "chunk_len" : 188,
    "chunk_right_context" : 1,
    "fc_d_model" : 512,
    "fifo_len" : 0,
    "max_index" : 99999,
    "min_pos_scores_rate" : 0.5,
    "model_type" : "sortformer_modules",
    "num_speakers" : 4,
    "pred_score_threshold" : 0.25,
    "scores_boost_latest" : 0.050000000000000003,
    "sil_threshold" : 0.20000000000000001,
    "spkcache_len" : 188,
    "spkcache_sil_frames_per_spk" : 3,
    "spkcache_update_period" : 188,
    "strong_boost_rate" : 0.75,
    "subsampling_factor" : 8,
    "tf_d_model" : 192,
    "use_aosc" : 1,
    "weak_boost_rate" : 1.5
  },
  "num_speakers" : 4,
  "pil_weight" : 0.5,
  "processor_config" : {
    "feature_size" : 128,
    "hop_length" : 160,
    "n_fft" : 512,
    "preemphasis" : 0.96999999999999997,
    "sampling_rate" : 16000,
    "win_length" : 400
  },
  "quantization" : {
    "bits" : 4,
    "group_size" : 64,
    "mode" : "affine"
  },
  "quantization_config" : {
    "bits" : 4,
    "group_size" : 64,
    "mode" : "affine"
  },
  "tf_encoder_config" : {
    "activation_function" : "relu",
    "d_model" : 192,
    "encoder_attention_heads" : 8,
    "encoder_ffn_dim" : 768,
    "encoder_layers" : 18,
    "k_proj_bias" : 1,
    "max_source_positions" : 1500,
    "model_type" : "sortformer_tf_encoder"
  }
}