mlc-chat-config.json

{
  "version": "0.1.0",
  "model_type": "deepseek_v2",
  "quantization": "q4f16_1",
  "model_config": {
    "vocab_size": 102400,
    "hidden_size": 2048,
    "intermediate_size": 10944,
    "moe_intermediate_size": 1408,
    "num_hidden_layers": 27,
    "num_attention_heads": 16,
    "num_key_value_heads": 16,
    "n_shared_experts": 2,
    "n_routed_experts": 64,
    "num_experts_per_tok": 6,
    "norm_topk_prob": false,
    "first_k_dense_replace": 1,
    "moe_layer_freq": 1,
    "routed_scaling_factor": 1.0,
    "attention_bias": false,
    "kv_lora_rank": 512,
    "qk_rope_head_dim": 64,
    "v_head_dim": 128,
    "qk_nope_head_dim": 128,
    "rms_norm_eps": 1e-06,
    "rope_theta": 10000,
    "q_lora_rank": null,
    "rope_scaling": {
      "beta_fast": 32,
      "beta_slow": 1,
      "factor": 40,
      "mscale": 0.707,
      "mscale_all_dim": 0.707,
      "original_max_position_embeddings": 4096,
      "type": "yarn"
    },
    "context_window_size": 163840,
    "prefill_chunk_size": 8192,
    "tensor_parallel_shards": 1,
    "dtype": "float32",
    "max_batch_size": 128
  },
  "vocab_size": 102400,
  "context_window_size": 163840,
  "sliding_window_size": -1,
  "prefill_chunk_size": 8192,
  "attention_sink_size": -1,
  "tensor_parallel_shards": 1,
  "pipeline_parallel_stages": 1,
  "temperature": 0.3,
  "presence_penalty": 0.0,
  "frequency_penalty": 0.0,
  "repetition_penalty": 1.0,
  "top_p": 0.95,
  "tokenizer_files": [
    "tokenizer.json",
    "tokenizer_config.json"
  ],
  "tokenizer_info": {
    "token_postproc_method": "byte_level",
    "prepend_space_in_encode": false,
    "strip_space_in_decode": false
  },
  "conv_template": {
    "name": "deepseek_v2",
    "system_template": "{system_message}",
    "system_message": "",
    "system_prefix_token_ids": [
      100000
    ],
    "add_role_after_system_message": true,
    "roles": {
      "user": "User",
      "assistant": "Assistant"
    },
    "role_templates": {
      "user": "{user_message}",
      "assistant": "{assistant_message}",
      "tool": "{tool_message}"
    },
    "messages": [],
    "seps": [
      "\n\n",
      "<\uff5cend\u2581of\u2581sentence\uff5c>"
    ],
    "role_content_sep": ": ",
    "role_empty_sep": ":",
    "stop_str": [
      "<\uff5cend\u2581of\u2581sentence\uff5c>"
    ],
    "stop_token_ids": [
      100001
    ],
    "function_string": "",
    "use_function_calling": false
  },
  "pad_token_id": 0,
  "bos_token_id": 100000,
  "eos_token_id": 100001
}