Fix EOS token: set token 50 as EOS

Fix EOS token: set token 50 (start_function_response) as EOS for proper stop

Add f16 GGUF

Add Q8_0 GGUF for llama.cpp

initial commit