config.json

{
    "architectures": [
      "Phi4FlashForCausalLM"
    ],
    "attention_dropout": 0.0,
    "auto_map": {
      "AutoConfig": "configuration_phi4flash.Phi4FlashConfig",
      "AutoModelForCausalLM": "modeling_phi4flash.Phi4FlashForCausalLM",
      "AutoTokenizer": "Xenova/gpt-4o"
    },
    "pad_token_id": 199999,
    "bos_token_id": 199999,
    "embd_pdrop": 0.0,
    "eos_token_id": 199999,
    "hidden_act": "silu",
    "hidden_size": 2560,
    "initializer_range": 0.02,
    "intermediate_size": 10240,
    "layer_norm_eps": 1e-5,
    "max_position_embeddings": 262144,
    "_attn_implementation": "flash_attention_2",
    "mb_per_layer": 2,
    "model_type": "phi4flash",
    "num_attention_heads": 40,
    "num_hidden_layers": 32,
    "num_key_value_heads": 20,
    "resid_pdrop": 0.0,
    "sliding_window": 512,
    "torch_dtype": "bfloat16",
    "tie_word_embeddings": true,
    "transformers_version": "4.46.1",
    "use_cache": true,
    "mlp_bias": false,
    "lm_head_bias": false,
    "vocab_size": 200064
  }