config.json

{
  "add_all_frames_to_correct_as_cond": false,
  "add_tpos_enc_to_obj_ptrs": false,
  "architectures": [
    "Sam2Model"
  ],
  "backbone_stride": 16,
  "binarize_mask_from_pts_for_mem_enc": false,
  "compile_image_encoder": false,
  "directly_add_no_mem_embed": true,
  "fixed_no_obj_ptr": true,
  "image_encoder_config": {
    "backbone_channel_list": [
      1152,
      576,
      288,
      144
    ],
    "embed_dim": 144,
    "global_att_blocks": [
      23,
      33,
      43
    ],
    "model_type": "",
    "stages": [
      2,
      6,
      36,
      4
    ],
    "window_pos_embed_bkg_spatial_size": [
      7,
      7
    ],
    "window_spec": [
      8,
      4,
      16,
      8
    ]
  },
  "image_size": 1024,
  "initializer_range": 0.02,
  "iou_prediction_use_sigmoid": true,
  "max_cond_frames_in_attn": -1,
  "max_obj_ptrs_in_encoder": 16,
  "memory_attention_config": {
    "model_type": ""
  },
  "memory_encoder_config": {
    "model_type": ""
  },
  "memory_temporal_stride_for_eval": 1,
  "model_type": "sam2",
  "multimask_max_pt_num": 1,
  "multimask_min_pt_num": 0,
  "multimask_output_for_tracking": true,
  "multimask_output_in_sam": true,
  "non_overlap_masks_for_mem_enc": false,
  "num_maskmem": 7,
  "only_obj_ptrs_in_the_past_for_eval": true,
  "pred_obj_scores": true,
  "pred_obj_scores_mlp": true,
  "proj_tpos_enc_in_obj_ptrs": false,
  "sam_mask_decoder_extra_args": null,
  "sigmoid_bias_for_mem_enc": -10,
  "sigmoid_scale_for_mem_enc": 20,
  "soft_no_obj_ptr": false,
  "torch_dtype": "float32",
  "transformers_version": "4.44.0.dev0",
  "use_high_res_features_in_sam": true,
  "use_mask_input_as_output_without_sam": true,
  "use_mlp_for_obj_ptr_proj": true,
  "use_multimask_token_for_obj_ptr": true,
  "use_obj_ptrs_in_encoder": true
}