Upload model

config.json

@@ -24,21 +24,130 @@
   "directly_add_no_memory_embedding": true,
   "fixed_no_object_pointer": true,
   "image_encoder_config": {
-    "model_type": ""
+    "backbone_channel_list": [
+      768,
+      384,
+      192,
+      96
+    ],
+    "dim_mul": 2.0,
+    "drop_path_rate": 0.0,
+    "fpn_hidden_size": 256,
+    "fpn_interpolation_mode": "nearest",
+    "fpn_kernel_size": 1,
+    "fpn_padding": 0,
+    "fpn_stride": 1,
+    "fpn_top_down_levels": [
+      2,
+      3
+    ],
+    "fuse_type": "sum",
+    "global_attention_blocks": [
+      5,
+      7,
+      9
+    ],
+    "head_mul": 2.0,
+    "hidden_act": "gelu",
+    "hidden_size": 96,
+    "image_size": 1024,
+    "layer_norm_eps": 1e-06,
+    "model_type": "",
+    "num_channels": 3,
+    "num_heads": 1,
+    "patch_kernel_size": 7,
+    "patch_padding": 3,
+    "patch_stride": 4,
+    "q_pool": 3,
+    "q_stride": [
+      2,
+      2
+    ],
+    "stages": [
+      1,
+      2,
+      7,
+      2
+    ],
+    "window_positional_embedding_background_size": [
+      7,
+      7
+    ],
+    "window_spec": [
+      8,
+      4,
+      14,
+      7
+    ]
   },
   "image_size": 1024,
   "initializer_range": 0.02,
   "iou_prediction_use_sigmoid": true,
   "mask_decoder_config": {
-    "model_type": ""
+    "dynamic_multimask_stability_delta": 0.05,
+    "dynamic_multimask_stability_thresh": 0.98,
+    "dynamic_multimask_via_stability": true,
+    "feed_forward_hidden_act": "relu",
+    "hidden_act": "gelu",
+    "hidden_size": 256,
+    "iou_head_depth": 3,
+    "iou_head_hidden_dim": 256,
+    "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,
   "max_object_pointers_in_encoder": 16,
   "memory_attention_config": {
-    "model_type": ""
+    "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",
+    "hidden_size": 256,
+    "model_type": "",
+    "num_layers": 4,
+    "rope_downsample_rate": 1,
+    "rope_dropout": 0.1,
+    "rope_embedding_dim": 256,
+    "rope_feat_sizes": [
+      32,
+      32
+    ],
+    "rope_num_heads": 1,
+    "rope_theta": 10000
   },
   "memory_encoder_config": {
-    "model_type": ""
+    "hidden_size": 256,
+    "mask_downsampler_embed_dim": 256,
+    "mask_downsampler_hidden_act": "gelu",
+    "mask_downsampler_kernel_size": 3,
+    "mask_downsampler_padding": 1,
+    "mask_downsampler_stride": 2,
+    "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,
+    "memory_fuser_padding": 3,
+    "memory_fuser_use_depthwise_conv": true,
+    "model_type": "",
+    "output_channels": 64
   },
   "memory_temporal_stride_for_eval": 1,
   "model_type": "sam2",
@@ -54,14 +163,22 @@
   "pred_obj_scores_mlp": true,
   "proj_tpos_enc_in_object_pointers": true,
   "prompt_encoder_config": {
-    "model_type": ""
+    "hidden_act": "gelu",
+    "hidden_size": 256,
+    "image_size": 1024,
+    "layer_norm_eps": 1e-06,
+    "mask_input_channels": 16,
+    "model_type": "",
+    "num_point_embeddings": 4,
+    "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,
   "torch_dtype": "float32",
-  "transformers_version": "4.47.0.dev0",
+  "transformers_version": "4.50.0.dev0",
   "use_mask_input_as_output_without_sam": true,
   "use_mlp_for_object_pointer_proj": true,
   "use_multimask_token_for_object_pointer": true,