fix keyboard_dim_in and remove dtype field

image_encoder/config.json

@@ -3,8 +3,6 @@
     "CLIPVisionModel"
   ],
   "attention_dropout": 0.0,
-  "dropout": 0.0,
-  "dtype": "float32",
   "hidden_act": "gelu",
   "hidden_size": 1280,
   "image_size": 224,

image_processor/preprocessor_config.json

@@ -24,4 +24,4 @@
   "size": {
     "shortest_edge": 224
   }
-}
+}

transformer/config.json

@@ -54,7 +54,7 @@
     "heads_num": 16,
     "hidden_size": 128,
     "img_hidden_size": 1536,
-    "keyboard_dim_in": 4,
+    "keyboard_dim_in": 6,
     "keyboard_hidden_dim": 1024,
     "mouse_dim_in": 2,
     "mouse_hidden_dim": 1024,