Add llama3 (#30334)

* nuke

* add co-author

* add co-author

* update card

* fixup and fix copies to please our ci

* nit fixup

* super small nits

* remove tokenizer_path from call to `write_model`

* always safe serialize by default

---------

Co-authored-by: pcuenca <pcuenca@users.noreply.github.com>
Co-authored-by: xenova <xenova@users.noreply.github.com>

utils/check_table.py

@@ -155,6 +155,7 @@ MODEL_NAMES_WITH_SAME_CONFIG = {
     "HerBERT": "BERT",
     "LayoutXLM": "LayoutLMv2",
     "Llama2": "LLaMA",
+    "Llama3": "LLaMA",
     "MADLAD-400": "T5",
     "MatCha": "Pix2Struct",
     "mBART-50": "mBART",