{
  "model_type": "phogpt",
  "architectures": ["PhoGPTForCausalLM"],
  "hidden_size": 4096,
  "num_attention_heads": 32,
  "num_hidden_layers": 32,
  "vocab_size": 51200,
  "max_seq_len": 4096,
  "torch_dtype": "float16",
  "attn_config": {
    "attn_impl": "triton",
    "attn_pdrop": 0.0,
    "attn_type": "multihead_attention",
    "alibi": true,
    "alibi_bias_max": 8,
    "prefix_lm": false,
    "qk_ln": false
  },
  "ffn_config": {
    "ffn_type": "phogpt_mlp"
  },
  "init_config": {
    "name": "kaiming_normal_",
    "fan_mode": "fan_in",
    "init_nonlinearity": "relu",
    "init_div_is_residual": true
  },
  "use_cache": false
}