Upload config.json with huggingface_hub

config.json

@@ -1,29 +1,18 @@
 {
   "model_type": "oculus",
   "architectures": ["OculusForConditionalGeneration"],
+  "architecture_name": "Oceanir-Oculus OO1",
   "version": "0.1",
 
-  "vision_encoders": {
-    "dinov3": {
-      "model_id": "facebook/dinov3-vitl16-pretrain-lvd1689m",
-      "hidden_size": 1024,
-      "num_layers": 24,
-      "num_heads": 16,
-      "image_size": 224,
-      "patch_size": 16
-    },
-    "siglip2": {
-      "model_id": "google/siglip2-so400m-patch16-naflex",
-      "hidden_size": 1152,
-      "num_layers": 27,
-      "num_heads": 16,
-      "image_size": 384,
-      "patch_size": 16
-    }
+  "vision_config": {
+    "hidden_size": 1024,
+    "num_layers": 24,
+    "num_heads": 16,
+    "image_size": 224,
+    "patch_size": 16
   },
 
-  "language_model": {
-    "model_id": "LiquidAI/LFM2.5-1.2B-Base",
+  "language_config": {
     "hidden_size": 1536,
     "num_layers": 16,
     "num_heads": 24,
@@ -69,7 +58,11 @@
   "structured_output": {
     "enabled": true,
     "json_token": "<json>",
-    "json_end_token": "</json>"
+    "json_end_token": "</json>",
+    "box_token": "<box>",
+    "box_end_token": "</box>",
+    "point_token": "<point>",
+    "point_end_token": "</point>"
   },
 
   "ocr": {