d7d2e6a0f0
* server: add option to output probabilities for completion * server: fix issue when handling probability output for incomplete tokens for multibyte character generation * server: fix llama_sample_top_k order * examples/common.h: put all bool variables in gpt_params together