config.json

{
  "name_or_path": "tclf90/KAT-V1-40B-GPTQ-Int4-Int8Mix",
  "architectures": [
    "Qwen2ForCausalLM"
  ],
  "attention_dropout": 0.0,
  "bos_token_id": 151643,
  "eos_token_id": 151643,
  "hidden_act": "silu",
  "hidden_size": 5120,
  "initializer_range": 0.02,
  "intermediate_size": 27648,
  "max_position_embeddings": 131072,
  "max_window_layers": 64,
  "model_type": "qwen2",
  "num_attention_heads": 40,
  "num_hidden_layers": 80,
  "num_key_value_heads": 8,
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "rope_theta": 1000000.0,
  "sliding_window": null,
  "tie_word_embeddings": false,
  "torch_dtype": "float16",
  "transformers_version": "4.46.1",
  "use_cache": false,
  "use_sliding_window": false,
  "vocab_size": 152064,
  "quantization_config": {
    "quant_method": "gptq",
    "bits": 4,
    "group_size": 128,
    "sym": false,
    "desc_act": false,
    "dynamic": {
      "+:model[.]layers[.]([0-5]|7[6-9])[.].*": {
        "bits": 8
      },
      "+:model[.]layers[.]([0-9]|7[0-9])[.]self_attn[.]o_proj.*": {
        "bits": 8
      },
      "+:model[.]layers[.]([0-9]|7[0-9])[.]mlp.down_proj.*": {
        "bits": 8
      }
    }
  }
}