config.json

{
  "model_type": "transformer",
  "hidden_size": 1024,
  "num_attention_heads": 16,
  "num_hidden_layers": 24,
  "intermediate_size": 4096,
  "hidden_act": "gelu",
  "initializer_range": 0.02,
  "layer_norm_eps": 1e-12,
  "max_position_embeddings": 512,
  "vocab_size": 50257,
  "type_vocab_size": 2,
  "pad_token_id": 0,
  "bos_token_id": 50256,
  "eos_token_id": 50256,
  "memory_module": {
    "short_term_memory_size": 128,
    "long_term_memory_size": 1024,
    "memory_update_method": "surprise",
    "forgetting_mechanism": "adaptive"
  }
}