config.json

{
  "architectures": [
    "OpenMythosForCausalLM"
  ],
  "bias": false,
  "block_size": 256,
  "bos_token_id": 1,
  "dtype": "float32",
  "effective_expected_depth": 10,
  "eos_token_id": 2,
  "head_dim": 64,
  "init_values": {
    "embed_scale": 16.0,
    "embedding": 0.03952847075210474,
    "out_proj": 0.008838834764831844,
    "std": 0.03952847075210474
  },
  "injection_type": "lti",
  "intermediate_size": 1024,
  "kv_lora_rank": 512,
  "mean_backprop_depth": 2,
  "mean_recurrence": 4,
  "model_type": "open_mythos",
  "moe_intermediate_size": 1024,
  "moe_top_k": 2,
  "n_embd": 256,
  "n_heads": 4,
  "n_kv_heads": 4,
  "n_layers": 4,
  "n_layers_in_coda": 1,
  "n_layers_in_prelude": 1,
  "n_layers_in_recurrent_block": 2,
  "n_routed_experts": 8,
  "n_shared_experts": 2,
  "norm_eps": 1e-06,
  "num_key_value_heads": 4,
  "pad_token_id": 0,
  "q_lora_rank": 1536,
  "qk_bias": true,
  "rope_base": 50000.0,
  "rope_head_dim": 64,
  "state_init": "like-init",
  "tie_embeddings": true,
  "tie_word_embeddings": true,
  "transformers_version": "5.6.0",
  "use_cache": false,
  "use_mla": false,
  "use_moe": false,
  "vocab_size": 50257
}