Add rf-detr-medium model

config.json

@@ -20,11 +20,10 @@
     "layer_norm_eps": 1e-06,
     "layerscale_value": 1.0,
     "mlp_ratio": 4,
-    "model_type": "rf_detr_dinov2_with_registers",
+    "model_type": "rf_detr_dinov2",
     "num_attention_heads": 6,
     "num_channels": 3,
     "num_hidden_layers": 12,
-    "num_register_tokens": 0,
     "num_windows": 2,
     "out_features": [
       "stage3",
@@ -56,6 +55,7 @@
       "stage11",
       "stage12"
     ],
+    "use_mask_token": true,
     "use_swiglu_ffn": false,
     "window_block_indexes": [
       0,
@@ -72,7 +72,6 @@
   "backbone_kwargs": null,
   "bbox_cost": 5,
   "bbox_loss_coefficient": 5,
-  "bbox_reparam": true,
   "class_cost": 2,
   "d_model": 256,
   "decoder_activation_function": "relu",
@@ -282,7 +281,6 @@
   "model_type": "rf_detr",
   "num_feature_levels": 1,
   "num_queries": 300,
-  "position_embedding_type": "sine",
   "projector_in_channels": [
     256
   ],
@@ -291,7 +289,6 @@
     1.0
   ],
   "transformers_version": "5.0.0.dev0",
-  "two_stage": true,
   "use_pretrained_backbone": false,
   "use_timm_backbone": false
 }