llama.cpp

Reenable tokenizer test for LLaMa

#3096

Merged