config.json

{
  "architectures": [
    "TSPModelForPretraining"
  ],
  "dropout_prob": 0.1,
  "embedding_size": 128,
  "hidden_size": 256,
  "intermediate_size": 1024,
  "max_sequence_length": 128,
  "model_type": "tsp",
  "num_attention_heads": 4,
  "num_hidden_layers": 12,
  "pad_token_id": 0,
  "position_embedding_type": "absolute",
  "torch_dtype": "float32",
  "transformers_version": "4.17.0",
  "vocab_size": 30522
}