config.json

{
  "model_type": "bt4",
  "architectures": ["BT4"],
  "embedding_size": 1024,
  "embedding_dense_sz": 512,
  "encoder_layers": 15,
  "encoder_d_model": 1024,
  "encoder_heads": 32,
  "encoder_dff": 1536,
  "dropout_rate": 0.0,
  "pol_embedding_size": 1024,
  "policy_d_model": 1024,
  "val_embedding_size": 128,
  "use_smolgen": true,
  "smol_hidden_channels": 32,
  "smol_hidden_sz": 256,
  "smol_gen_sz": 256,
  "smol_activation": "swish",
  "auto_map": {
    "AutoConfig": "model.BT4Config",
    "AutoModel": "model.BT4"
  },
  "torch_dtype": "float32"
}