{
  "architectures": [
    "CustomTransformerForCausalLM"
  ],
  "d_model": 512,
  "dim_feedforward": 2048,
  "dropout": 0.1,
  "group_size": 16,
  "model_type": "custom_transformer",
  "n_heads": 8,
  "num_layers": 6,
  "torch_dtype": "float32",
  "transformers_version": "4.42.4",
  "vocab_size": 32000
}