config.json

{
  "_name_or_path": "lmsys/vicuna-7b-v1.5",
  "architectures": [
    "PmodLlavaLlamaForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "bos_token_id": 1,
  "eos_token_id": 2,
  "freeze_mm_mlp_adapter": false,
  "hidden_act": "silu",
  "hidden_size": 4096,
  "image_aspect_ratio": "anyres",
  "image_grid_pinpoints": [
    [
      336,
      672
    ],
    [
      672,
      336
    ],
    [
      672,
      672
    ],
    [
      1008,
      336
    ],
    [
      336,
      1008
    ]
  ],
  "initializer_range": 0.02,
  "intermediate_size": 11008,
  "max_num_patches": null,
  "max_position_embeddings": 4096,
  "mm_hidden_size": 1024,
  "mm_patch_merge_type": "spatial_unpad",
  "mm_projector_lr": null,
  "mm_projector_type": "mlp2x_gelu",
  "mm_use_im_patch_token": false,
  "mm_use_im_start_end": false,
  "mm_vision_select_feature": "patch",
  "mm_vision_select_layer": -2,
  "mm_vision_tower": "openai/clip-vit-large-patch14-336",
  "mm_vision_tower_lr": null,
  "mod_bias_enabled": true,
  "mod_router_factor": 0.5,
  "mod_special_init": true,
  "mod_target_layer_indices": [
    8,
    9,
    10,
    11,
    12,
    13,
    14,
    15,
    16,
    17,
    18,
    19,
    20,
    21,
    22,
    23,
    24,
    25,
    26,
    27,
    28,
    29,
    30,
    31
  ],
  "mod_target_layers": "shiftedcos_decay_0.85_0.15",
  "mod_target_token_type": "vision_all",
  "mod_weight_norm": "tanh_0.2_1",
  "model_type": "pmod_llava_llama",
  "num_attention_heads": 32,
  "num_hidden_layers": 32,
  "num_key_value_heads": 32,
  "pad_token_id": 0,
  "pretraining_tp": 1,
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "rope_theta": 10000.0,
  "tie_word_embeddings": false,
  "tokenizer_model_max_length": 4096,
  "tokenizer_padding_side": "right",
  "torch_dtype": "bfloat16",
  "transformers_version": "4.37.2",
  "tune_mm_mlp_adapter": false,
  "unfreeze_mm_vision_tower": true,
  "use_cache": true,
  "use_mm_proj": true,
  "vocab_size": 32000
}