{
  "vocab_size": 4096,
  "n_layers": 6,
  "d_model": 256,
  "n_heads": 4,
  "context": 512,
  "dropout": 0.1,
  "attn_dropout": 0.0,
  "architecture": "decoder-only transformer (RoPE + RMSNorm + SwiGLU)",
  "tie_word_embeddings": true,
  "trained_epoch": 12,
  "train_loss": 3.5534612417221068,
  "val_loss": 4.055904150009155,
  "params": 5868800
}