Update config.json

Fix a bug resulting from the transformers/models/llama/convert_llama_weights_to_hf.py.

config.json

@@ -8,7 +8,7 @@
   "hidden_size": 4096,
   "initializer_range": 0.02,
   "intermediate_size": 11008,
-  "max_position_embeddings": 2048,
+  "max_position_embeddings": 4096,
   "model_type": "llama",
   "num_attention_heads": 32,
   "num_hidden_layers": 32,