config.json

{
  "_bb_feat_sizes": [
    [
      256,
      256
    ],
    [
      128,
      128
    ],
    [
      64,
      64
    ]
  ],
  "add_all_frames_to_correct_as_cond": false,
  "add_tpos_enc_to_object_pointers": false,
  "architectures": [
    "Sam2Model"
  ],
  "backbone_stride": 16,
  "binarize_mask_from_pts_for_mem_enc": false,
  "compile_image_encoder": false,
  "directly_add_no_memory_embedding": true,
  "fixed_no_object_pointer": true,
  "image_encoder_config": {
    "model_type": ""
  },
  "image_size": 1024,
  "initializer_range": 0.02,
  "iou_prediction_use_sigmoid": true,
  "mask_decoder_config": {
    "model_type": ""
  },
  "max_cond_frames_in_attn": -1,
  "max_object_pointers_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,
  "no_obj_embed_spatial": true,
  "non_overlap_masks_for_mem_enc": false,
  "num_maskmem": 7,
  "only_object_pointers_in_the_past_for_eval": true,
  "pred_obj_scores": true,
  "pred_obj_scores_mlp": true,
  "proj_tpos_enc_in_object_pointers": true,
  "prompt_encoder_config": {
    "model_type": ""
  },
  "sam_mask_decoder_extra_args": null,
  "sigmoid_bias_for_mem_enc": -10,
  "sigmoid_scale_for_mem_enc": 20,
  "soft_no_object_pointer": false,
  "torch_dtype": "float32",
  "transformers_version": "4.47.0.dev0",
  "use_mask_input_as_output_without_sam": true,
  "use_mlp_for_object_pointer_proj": true,
  "use_multimask_token_for_object_pointer": true,
  "use_object_pointers_in_encoder": true,
  "use_signed_tpos_enc_to_object_pointers": true
}