model_config.json

{
  "vocab_size": 32000,
  "block_size": 512,
  "n_embd": 512,
  "n_head": 8,
  "n_layer": 8,
  "use_rope": true,
  "use_rmsnorm": true,
  "use_qk_norm": true,
  "use_relu2": true,
  "logit_cap": 15.0,
  "use_chunked_loss": true,
  "use_zero_init": true
}