llama.cpp

Update special token handling in conversion scripts for gpt2 derived tokenizers

#3746

Merged