config.json

{
  "architectures": [
    "Qwen3ForCausalLM"
  ],
  "model_type": "qwen3",
  "vocab_size": 259,
  "hidden_size": 512,
  "num_hidden_layers": 16,
  "intermediate_size": 1536,
  "num_attention_heads": 4,
  "num_key_value_heads": 2,
  "head_dim": 128,
  "hidden_act": "silu",
  "rms_norm_eps": 1e-06,
  "max_position_embeddings": 4096,
  "rope_theta": 1000000.0,
  "rope_scaling": null,
  "attention_bias": false,
  "attention_dropout": 0.0,
  "tie_word_embeddings": false,
  "use_cache": true,
  "bos_token_id": 256,
  "eos_token_id": 257,
  "pad_token_id": 258,
  "torch_dtype": "float32"
}