config.json

{
  "vocab_size": 32768,
  "max_seq_len": 4096,
  "n_layer": 32,
  "n_embd": 576,
  "n_head": 9,
  "n_kv_head": 3,
  "mlp_hidden": 1536,
  "rope_base": 10000.0,
  "rms_eps": 1e-06,
  "qk_norm": true,
  "z_loss": 0.0,

  "model_type": "gouda-gouda",
  "architectures": [
    "GoudaForCausalLM"
  ],

  "auto_map": {
    "AutoConfig": "configuration_gouda.GoudaConfig",
    "AutoModelForCausalLM": "modeling_gouda.GoudaForCausalLM"
  },

  "name": "Gouda-2.0",
  "torch_dtype": "bfloat16",
  "tie_word_embeddings": true
}