config.json

{
  "_name_or_path": "llama3_2_1b",
  "architectures": [
    "LlamaForCausalLM"
  ],
  "asym": false,
  "attention_bias": false,
  "attention_dropout": 0.0,
  "bigdl_transformers_low_bit": "sym_int4_rtn",
  "bos_token_id": 128000,
  "const_parameter": true,
  "cos_sin_input": true,
  "embedding_post": true,
  "eos_token_id": [
    128001,
    128008,
    128009
  ],
  "fused_layers": 2,
  "group_size": 0,
  "head_dim": 64,
  "hidden_act": "silu",
  "hidden_size": 2048,
  "initializer_range": 0.02,
  "intermediate_size": 8192,
  "kv_len": 1023,
  "lm_head_low_bit": "sym_int4_rtn",
  "max_position_embeddings": 131072,
  "max_prompt_len": 512,
  "mixed_precision": false,
  "mlp_bias": false,
  "model_type": "llama",
  "n_splits_down_proj": 1,
  "n_splits_linear": 1,
  "num_attention_heads": 32,
  "num_head": 32,
  "num_hidden_layers": 16,
  "num_key_value_heads": 8,
  "optimize_model": true,
  "pretraining_tp": 1,
  "qkv_bias": false,
  "rms_norm_eps": 1e-05,
  "rope_scaling": {
    "factor": 32.0,
    "high_freq_factor": 4.0,
    "low_freq_factor": 1.0,
    "original_max_position_embeddings": 8192,
    "rope_type": "llama3"
  },
  "rope_theta": 500000.0,
  "tie_word_embeddings": true,
  "torch_dtype": "float16",
  "transformers_version": "4.45.0",
  "transpose_value_cache": true,
  "use_cache": true,
  "use_prefill_sdp": false,
  "vocab_size": 128256,
  "weight_idx": 5,
  "weight_num": 7
}