config.json

{
  "architectures": [
    "InternLM3ForCausalLM"
  ],
  "attention_dropout": 0.0,
  "auto_map": {
    "AutoConfig": "internlm/internlm3-8b-instruct--configuration_internlm3.InternLM3Config",
    "AutoModel": "internlm/internlm3-8b-instruct--modeling_internlm3.InternLM3Model",
    "AutoModelForCausalLM": "modeling_internlm3.InternLM3ForCausalLM"
  },
  "bias": false,
  "bos_token_id": 1,
  "eos_token_id": 2,
  "head_dim": 128,
  "hidden_act": "silu",
  "hidden_size": 4096,
  "initializer_range": 0.02,
  "intermediate_size": 10240,
  "max_position_embeddings": 32768,
  "model_type": "internlm3",
  "num_attention_heads": 32,
  "num_hidden_layers": 48,
  "num_key_value_heads": 2,
  "pad_token_id": 2,
  "qkv_bias": false,
  "quantization_config": {
    "act_bits": 16,
    "act_data_type": "int",
    "act_dynamic": true,
    "act_group_size": 64,
    "act_sym": false,
    "amp": false,
    "autoround_version": "0.4.7",
    "batch_size": 4,
    "bits": 4,
    "data_type": "int",
    "dataset": "NeelNanda/pile-10k",
    "enable_minmax_tuning": true,
    "enable_norm_bias_tuning": false,
    "enable_quanted_input": true,
    "gradient_accumulate_steps": 1,
    "group_size": 64,
    "iters": 200,
    "low_gpu_mem_usage": false,
    "lr": 0.005,
    "minmax_lr": 0.005,
    "modules_to_not_convert": [
      "lm_head"
    ],
    "nsamples": 128,
    "quant_method": "awq",
    "scale_dtype": "torch.float16",
    "seqlen": 512,
    "super_bits": null,
    "super_group_size": null,
    "sym": false,
    "to_quant_block_names": null,
    "version": "gemm",
    "zero_point": true
  },
  "rms_norm_eps": 1e-05,
  "rope_scaling": {
    "factor": 6.0,
    "rope_type": "dynamic"
  },
  "rope_theta": 50000000,
  "tie_word_embeddings": false,
  "torch_dtype": "float16",
  "transformers_version": "4.51.3",
  "use_cache": true,
  "vocab_size": 128512
}