config.json

{
  "context_length": 2048,
  "dropout": 0.0,
  "embedding_dimensions": 512,
  "hidden_ratio": 2,
  "kv_heads": 4,
  "num_encoder_layers": 4,
  "padding_index": 1,
  "q_heads": 16,
  "teacher_dimensions": 1152,
  "vocabulary_size": 33
}