{
  "architectures": [
    "T5EncoderModel"
  ],
  "model_type": "t5",
  "d_model": 1024,
  "d_kv": 64,
  "d_ff": 4096,
  "num_heads": 16,
  "num_layers": 24,
  "vocab_size": 32128,
  "relative_attention_num_buckets": 32,
  "relative_attention_max_distance": 128,
  "dropout_rate": 0.0,
  "layer_norm_epsilon": 1e-06,
  "feed_forward_proj": "relu",
  "tie_word_embeddings": true,
  "decoder_start_token_id": 0
}