config.json

{
  "architectures": ["GPT2LMHeadModel"],
  "model_type": "gpt2",
  "model_name": "OMDA",
  "architecture": "OMDA-Decoder",
  "vocab_size": 128004,
  "d_model": 512,
  "n_layers": 6,
  "n_heads": 8,
  "d_ff": 2048,
  "max_seq_len": 512,
  "dropout": 0.1,
  "batch_size": 8,
  "learning_rate": 0.0001,
  "num_epochs": 5,
  "save_steps": 1000,
  "eval_steps": 500,
  "model_save_path": "./omda_chat_model",
  "tokenizer_name": "OMDATokenizer"
}