Upload config.json with huggingface_hub

config.json

@@ -39,16 +39,53 @@
   "projector_config": {
     "fused_vision_dim": 2176,
     "hidden_dim": 4352,
-    "output_dim": 1536
+    "output_dim": 1536,
+    "num_tokens": 64
   },
 
   "task_heads": {
     "segmentation_classes": 150,
-    "detection_classes": 80
+    "detection_classes": 80,
+    "ui_element_classes": 50
   },
 
-  "reasoning_enabled": true,
-  "thinking_token": "<think>",
-  "thinking_end_token": "</think>",
-  "max_thinking_tokens": 256
+  "reasoning": {
+    "enabled": true,
+    "thinking_token": "<think>",
+    "thinking_end_token": "</think>",
+    "max_thinking_tokens": 256,
+    "style": "structured"
+  },
+
+  "focus_system": {
+    "enabled": true,
+    "focus_token": "<focus>",
+    "focus_end_token": "</focus>",
+    "max_regions": 4,
+    "min_crop_size": 64,
+    "auto_threshold": 0.7
+  },
+
+  "structured_output": {
+    "enabled": true,
+    "json_token": "<json>",
+    "json_end_token": "</json>"
+  },
+
+  "ocr": {
+    "enabled": true,
+    "languages": ["en"],
+    "confidence_threshold": 0.5
+  },
+
+  "ui_understanding": {
+    "enabled": true,
+    "element_classes": 50
+  },
+
+  "generation": {
+    "max_new_tokens": 512,
+    "temperature": 0.7,
+    "top_p": 0.95
+  }
 }