Training in progress, epoch 1

config.json

@@ -1,9 +1,9 @@
 {
-  "_name_or_path": "facebook/detr-resnet-50",
+  "_name_or_path": "microsoft/conditional-detr-resnet-50",
   "activation_dropout": 0.0,
   "activation_function": "relu",
   "architectures": [
-    "DetrForObjectDetection"
+    "ConditionalDetrForObjectDetection"
   ],
   "attention_dropout": 0.0,
   "auxiliary_loss": false,
@@ -20,8 +20,8 @@
   },
   "bbox_cost": 5,
   "bbox_loss_coefficient": 5,
-  "class_cost": 1,
-  "classifier_dropout": 0.0,
+  "class_cost": 2,
+  "cls_loss_coefficient": 2,
   "d_model": 256,
   "decoder_attention_heads": 8,
   "decoder_ffn_dim": 2048,
@@ -34,7 +34,7 @@
   "encoder_ffn_dim": 2048,
   "encoder_layerdrop": 0.0,
   "encoder_layers": 6,
-  "eos_coefficient": 0.1,
+  "focal_alpha": 0.25,
   "giou_cost": 2,
   "giou_loss_coefficient": 2,
   "id2label": {
@@ -50,10 +50,10 @@
   },
   "mask_loss_coefficient": 1,
   "max_position_embeddings": 1024,
-  "model_type": "detr",
+  "model_type": "conditional_detr",
   "num_channels": 3,
   "num_hidden_layers": 6,
-  "num_queries": 100,
+  "num_queries": 300,
   "position_embedding_type": "sine",
   "scale_embedding": false,
   "torch_dtype": "float32",

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:db8147c64bf0fe7c64c0e49ec7b6e0d6e99ac79770094b33983a2f4e372917cb
-size 166495852
+oid sha256:f23147b760a58a75055bb9bcc5c97ad63f0d7baf50171800a6fa83517a983bd7
+size 174076712

preprocessor_config.json

@@ -10,7 +10,7 @@
     0.456,
     0.406
   ],
-  "image_processor_type": "DetrImageProcessor",
+  "image_processor_type": "ConditionalDetrImageProcessor",
   "image_std": [
     0.229,
     0.224,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:421a3d3a351f0b82ab9c0c4f99a6abd1a43ded95359019e9043cb01160990a5d
+oid sha256:c675ecbf859926ce2167a3dff047e2f2d4a3e1306306fba129b07dec7b511764
 size 5112