Set `tie_word_embeddings` to False

Fixes the warning below when using multiple GPUs. The parameter is not utilised by Llava (and set to `false` for Llama)

```
2023-12-14 19:03:27,491 - accelerate.utils.modeling - WARNING - The model weights are not tied. Please use the `tie_weights` method before using the `infer_auto_device` function.
```

config.json

@@ -18,6 +18,7 @@
     "torch_dtype": "float16",
     "vocab_size": 32064
   },
+  "tie_word_embeddings": false,
   "torch_dtype": "float16",
   "transformers_version": "4.36.0.dev0",
   "vision_config": {