config.json

{
  "architectures": [
    "Phi3ForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "bos_token_id": 100257,
  "dfloat11_config": {
    "bytes_per_thread": 8,
    "pattern_dict": {
      "lm_head": [],
      "model.embed_tokens": [],
      "model.layers.\\d+": [
        "self_attn.o_proj",
        "self_attn.qkv_proj",
        "mlp.gate_up_proj",
        "mlp.down_proj"
      ]
    },
    "threads_per_block": [
      512
    ],
    "version": "0.2.0"
  },
  "embd_pdrop": 0.0,
  "eos_token_id": 100265,
  "hidden_act": "silu",
  "hidden_size": 5120,
  "initializer_range": 0.02,
  "intermediate_size": 17920,
  "max_position_embeddings": 32768,
  "model_type": "phi3",
  "num_attention_heads": 40,
  "num_hidden_layers": 40,
  "num_key_value_heads": 10,
  "original_max_position_embeddings": 32768,
  "pad_token_id": 100349,
  "partial_rotary_factor": 1.0,
  "resid_pdrop": 0.0,
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "rope_theta": 500000,
  "sliding_window": null,
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.51.3",
  "use_cache": true,
  "vocab_size": 100352
}