config.json

{
  "_name_or_path": "ShapeLLM_13B_general_v1.0",
  "architectures": [
    "LlavaLlamaForCausalLM"
  ],
  "bos_token_id": 1,
  "eos_token_id": 2,
  "freeze_mm_mlp_adapter": false,
  "hidden_act": "silu",
  "hidden_size": 5120,
  "initializer_range": 0.02,
  "intermediate_size": 13824,
  "max_position_embeddings": 2048,
  "mm_hidden_size": 1024,
  "mm_projector_lr": null,
  "mm_projector_type": "mlp2x_gelu",
  "mm_use_pt_patch_token": false,
  "mm_use_pt_start_end": false,
  "mm_vision_select_feature": "patch",
  "mm_vision_select_layer": -2,
  "mm_vision_tower": "ReConV2/cfgs/pretrain/large/openshape.yaml",
  "model_type": "llava",
  "num_attention_heads": 40,
  "num_hidden_layers": 40,
  "num_key_value_heads": 40,
  "pad_token_id": 0,
  "pretraining_tp": 1,
  "prompt_token_num": 32,
  "rms_norm_eps": 1e-06,
  "rope_scaling": null,
  "sample_points_num": 10000,
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.31.0",
  "tune_mm_mlp_adapter": false,
  "use_cache": true,
  "use_mm_proj": true,
  "vision_tower_path": "./checkpoints/recon/large.pth",
  "vocab_size": 32000,
  "with_ape": true,
  "with_color": true,
  "with_global": true,
  "with_local": true
}