config.json

{
  "architectures": [
    "CustomStudentModel"
  ],
  "dropout": 0.1,
  "hidden_size": 512,
  "intermediate_size": 2048,
  "num_attention_heads": 8,
  "num_decoder_layers": 6,
  "num_encoder_layers": 6,
  "torch_dtype": "float32",
  "transformers_version": "4.47.0",
  "vocab_size": 256204
}