Fix tokenizer_config.json

Without this fix, if you do:
```py
from transformers import pipeline
pipeline('text-generation', 'JackFram/llama-160m')
```

you get an error:
```
RecursionError: maximum recursion depth exceeded while calling a Python object
```

See [this issue](https://github.com/huggingface/transformers/issues/25036) for more info.

tokenizer_config.json

@@ -3,7 +3,7 @@
   "add_eos_token": false,
   "bos_token": {
     "__type": "AddedToken",
-    "content": "",
+    "content": "<s>",
     "lstrip": false,
     "normalized": true,
     "rstrip": false,
@@ -12,7 +12,7 @@
   "clean_up_tokenization_spaces": false,
   "eos_token": {
     "__type": "AddedToken",
-    "content": "",
+    "content": "</s>",
     "lstrip": false,
     "normalized": true,
     "rstrip": false,
@@ -24,7 +24,7 @@
   "tokenizer_class": "LlamaTokenizer",
   "unk_token": {
     "__type": "AddedToken",
-    "content": "",
+    "content": "<unk>",
     "lstrip": false,
     "normalized": true,
     "rstrip": false,