llama.cpp

Respect tokenizer.ggml.add_bos_token value when tokenizing

#4040

Merged