model/model.json

{
  "max_seq_len": 1024,
  "d_model": 768,
  "n_layers": 12,
  "n_heads": 12,
  "use_padded_vocab_size": true,
  "use_rope": true,
  "rope_theta": 10000.0,
  "is_rope_full_precision": true,
  "embd_dropout_p": 0.0,
  "attn_dropout_p": 0.0,
  "residual_dropout_p": 0.0,
  "init_std": 0.02,
  "init_residual_scaled_factor": 2.0
}