Update config.json

config.json

@@ -5,7 +5,7 @@
   "alpha_type": "float",
   "alphas_initializer_range": 0.0,
   "architectures": [
-    "VLlamaForCausalLM"
+    "IdeficsForCausalLM"
   ],
   "bos_token_id": 1,
   "cross_layer_activation_function": "swiglu",
@@ -23,7 +23,7 @@
   "intermediate_size": 11008,
   "max_position_embeddings": 2048,
   "max_sequence_length": 2048,
-  "model_type": "vllama",
+  "model_type": "idefics",
   "num_attention_heads": 32,
   "num_hidden_layers": 32,
   "pad_token_id": 0,
@@ -41,7 +41,9 @@
   "use_resampler": true,
   "vision_embed_dim": 1280,
   "vision_image_size": 224,
-  "vision_model_name": "laion/CLIP-ViT-H-14-laion2B-s32B-b79K",
-  "vision_model_params": "{\"id2label\":{}, \"label2id\":{}}",
+  "vision_intermediate_size": 5120,
+  "vision_patch_size": 14,
+  "vision_num_attention_heads": 16,
+  "vision_num_hidden_layers": 32,
   "vocab_size": 32000
 }