config.json

{
  "model_type": "nvmos_text_query_scorer",
  "audio_encoder": "marcoyang/spear-large-speech-audio",
  "audio_layer": 9,
  "text_encoder": "FacebookAI/xlm-roberta-large",
  "audio_dim": 1024,
  "text_dim": 1024,
  "hidden_dim": 256,
  "attention_heads": 8,
  "cross_layers": 2,
  "ffn_dim": 1024,
  "dropout": 0.1,
  "score_range": [
    0,
    5
  ],
  "sample_rate": 16000,
  "max_audio_sec": 12.0,
  "max_text_len": 192,
  "query_mode": "tag_ctx"
}