Upload model

config.json

@@ -1,28 +1,13 @@
 {
-  "_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,
+  "binarize_mask_from_pts_for_mem_enc": true,
+  "enable_occlusion_spatial_embedding": true,
+  "enable_temporal_pos_encoding_for_object_pointers": true,
+  "fill_hole_area": 8,
   "image_encoder_config": {
     "backbone_channel_list": [
       768,
@@ -30,6 +15,20 @@
       192,
       96
     ],
+    "backbone_feature_sizes": [
+      [
+        256,
+        256
+      ],
+      [
+        128,
+        128
+      ],
+      [
+        64,
+        64
+      ]
+    ],
     "dim_mul": 2.0,
     "drop_path_rate": 0.0,
     "fpn_hidden_size": 256,
@@ -54,6 +53,7 @@
     "layer_norm_eps": 1e-06,
     "model_type": "",
     "num_channels": 3,
+    "num_feature_levels": 3,
     "num_heads": 1,
     "patch_kernel_size": 7,
     "patch_padding": 3,
@@ -95,15 +95,12 @@
     "iou_prediction_use_sigmoid": true,
     "model_type": "",
     "num_multimask_outputs": 3,
-    "pred_obj_scores": true,
-    "pred_obj_scores_mlp": true,
     "two_way_transformer_activation": "relu",
     "two_way_transformer_attention_downsample_rate": 2,
     "two_way_transformer_depth": 2,
     "two_way_transformer_embedding_dim": 256,
     "two_way_transformer_mlp_dim": 2048,
     "two_way_transformer_num_heads": 8,
-    "use_high_resolution_features": true,
     "use_multimask_token_for_object_pointer": true
   },
   "max_cond_frames_in_attn": -1,
@@ -111,9 +108,7 @@
   "memory_attention_config": {
     "apply_pe_at_cross_attn_keys": true,
     "apply_pe_at_cross_attn_queries": false,
-    "apply_pe_at_input": true,
     "apply_pe_at_self_attn": false,
-    "batch_first": true,
     "dim_feedforward": 2048,
     "dropout": 0.1,
     "hidden_act": "relu",
@@ -140,7 +135,6 @@
     "mask_downsampler_total_stride": 16,
     "memory_fuser_embed_dim": 256,
     "memory_fuser_hidden_act": "gelu",
-    "memory_fuser_input_projection": false,
     "memory_fuser_kernel_size": 7,
     "memory_fuser_layer_scale_init_value": 1e-06,
     "memory_fuser_num_layers": 2,
@@ -155,13 +149,11 @@
   "multimask_min_pt_num": 0,
   "multimask_output_for_tracking": true,
   "multimask_output_in_sam": true,
-  "no_obj_embed_spatial": true,
+  "non_overlap_masks": false,
   "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,
+  "preserve_temporal_direction_in_object_pointers": true,
+  "project_temporal_pos_encoding_in_object_pointers": true,
   "prompt_encoder_config": {
     "hidden_act": "gelu",
     "hidden_size": 256,
@@ -173,15 +165,11 @@
     "patch_size": 16,
     "scale": 1
   },
-  "sam_mask_decoder_extra_args": null,
-  "sigmoid_bias_for_mem_enc": -10,
-  "sigmoid_scale_for_mem_enc": 20,
-  "soft_no_object_pointer": false,
+  "sigmoid_bias_for_mem_enc": -10.0,
+  "sigmoid_scale_for_mem_enc": 20.0,
   "torch_dtype": "float32",
-  "transformers_version": "4.50.0.dev0",
+  "transformers_version": "4.53.0.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
+  "use_object_pointers_in_encoder": true
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c5644fa49b4c724cb7ff6fe267e31484dc27f64a28d15f26800ca6335d4be640
-size 155906128
+oid sha256:030a362dae8f614f7aa8afc76097ecacd3a22b00baf76f8f1d92819a5eedf3f0
+size 155906184