config.json

{
  "_name_or_path": "/mnt/projects4jw/jiteshjain_sherlock/models/ABLATE/seed42/SUBMISSION/d8-20_s10-18_g12-20_lossd0.5_s0.5_g0.5_textTrue_sherlock_gen-depth-seg_pretrain_CLIP-convnext_xxlarge-laion2B-s34B-b82K-augreg-soup-res768_llama3-8b_dsg_contw_0.3_contrastiveTrue_ceFalse_oneformer_attn_causal_n8",
  "architectures": [
    "OlaLlavaLlamaForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "attn_mask_type": "causal",
  "aux_mode": "gen-depth-seg",
  "bos_token_id": 128000,
  "contrastive_loss_weight": 0.3,
  "depth_estimator": "/mnt/projects4jw/jiteshjain_sherlock/depth_anything_v2_vitl.pth",
  "eos_token_id": 128009,
  "freeze_mm_mlp_adapter": false,
  "hidden_act": "silu",
  "hidden_size": 4096,
  "image_aspect_ratio": "square",
  "image_depth": {
    "depth": 1,
    "depth_layer_indices": "8-20",
    "depth_loss_weight": 0.5,
    "dim_head": 32,
    "ff_mult": 1,
    "num_heads": 4,
    "num_tokens": 576,
    "output_dim": 1024,
    "use_intermediate_depth": false
  },
  "image_gen": {
    "depth": 1,
    "dim_head": 32,
    "ff_mult": 1,
    "img_layer_indices": "12-20",
    "img_loss_weight": 0.5,
    "num_heads": 4,
    "num_tokens": 1,
    "output_dim": 1024
  },
  "image_generator": "/mnt/projects4jw/jiteshjain_sherlock/stable-diffusion-2-1-unclip",
  "image_grid_pinpoints": [
    [
      336,
      672
    ],
    [
      672,
      336
    ],
    [
      672,
      672
    ],
    [
      1008,
      336
    ],
    [
      336,
      1008
    ]
  ],
  "image_seg": {
    "depth": 1,
    "dim_head": 32,
    "ff_mult": 1,
    "num_heads": 4,
    "num_tokens": 576,
    "output_dim": 1536,
    "seg_layer_indices": "10-18",
    "seg_loss_weight": 0.5,
    "seg_teacher": "oneformer"
  },
  "image_segmentor": "/mnt/projects4jw/jiteshjain_sherlock/oneformer_coco_swin_large",
  "initializer_range": 0.02,
  "intermediate_size": 14336,
  "max_position_embeddings": 8192,
  "mlp_bias": false,
  "mm_hidden_size": 3072,
  "mm_patch_merge_type": "flat",
  "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": "/mnt/projects4jw/jiteshjain_sherlock/CLIP-convnext_xxlarge-laion2B-s34B-b82K-augreg-soup-res768",
  "model_type": "ola_llama",
  "num_attention_heads": 32,
  "num_hidden_layers": 32,
  "num_key_value_heads": 8,
  "num_task_tokens": 8,
  "pass_text_to_aux": true,
  "pretraining_tp": 1,
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "rope_theta": 500000.0,
  "s2_scales": "336,1008",
  "sample_tokens": false,
  "task_token_format": "expand_emb",
  "tie_word_embeddings": false,
  "tokenizer_model_max_length": 4096,
  "tokenizer_padding_side": "right",
  "torch_dtype": "float32",
  "transformers_version": "4.41.1",
  "tune_mm_mlp_adapter": true,
  "use_cache": true,
  "use_ce": false,
  "use_contrastive": true,
  "use_mm_proj": true,
  "use_s2": false,
  "vocab_size": 128257
}