config.json

{
  "_name_or_path": "./qg_model_multiscale_stage2",
  "architectures": [
    "PixelLMForCausalLM"
  ],
  "bce_loss_weight": 2.0,
  "bos_token_id": 1,
  "ce_loss_weight": 1.0,
  "dice_loss_weight": 0.5,
  "eos_token_id": 2,
  "freeze_mm_mlp_adapter": true,
  "freeze_mm_vision_resampler": false,
  "hidden_act": "silu",
  "hidden_size": 4096,
  "image_aspect_ratio": "square",
  "image_feature_scale_num": 2,
  "image_grid_pinpoints": null,
  "initializer_range": 0.02,
  "intermediate_size": 11008,
  "masks_process_with_clip": false,
  "max_length": 4096,
  "max_position_embeddings": 4096,
  "mm_hidden_size": 1024,
  "mm_projector_hidden_dim": 2,
  "mm_projector_out_dim": 1,
  "mm_projector_type": "mlp2x_gelu",
  "mm_resampler_type": null,
  "mm_use_im_patch_token": false,
  "mm_use_im_start_end": false,
  "mm_vision_select_feature": "patch",
  "mm_vision_select_layer": [
    6,
    -11,
    -2
  ],
  "mm_vision_tower": "openai/clip-vit-large-patch14-336",
  "model_type": "llava",
  "num_attention_heads": 32,
  "num_hidden_layers": 32,
  "num_key_value_heads": 32,
  "out_dim": 256,
  "pad_token_id": 0,
  "pad_train_clip_images": true,
  "pretrain_mm_mlp_adapter": null,
  "pretraining_tp": 1,
  "resize_vision_tower": true,
  "resize_vision_tower_size": 448,
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "seg_token_idx": [
    32000,
    32001,
    32002,
    32003,
    32004,
    32005
  ],
  "seg_token_num": 3,
  "separate_mm_projector": true,
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "train_mask_decoder": true,
  "transformers_version": "4.31.0",
  "tune_mm_mlp_adapter": false,
  "tune_mm_vision_resampler": false,
  "unfreeze_mm_vision_tower": false,
  "use_cache": false,
  "use_mm_proj": true,
  "use_mm_start_end": true,
  "use_multi_scale_mm_projector": true,
  "vision_pretrained": "PATH_TO_SAM_ViT-H",
  "vision_tower": "openai/clip-vit-large-patch14-336",
  "vision_tower_for_mask": true,
  "vocab_size": 32008
}