.\models\server\llama\llama-server -m .\models\llama\7b_model.gguf --port 30000 -c 2048 -n -1 -t 6 -ngl 256 --ignore-eos --api-key r6BvC8V3RkO5nJe93i1KMHiqaDi6ZWaM --override-kv tokenizer.ggml.add_bos_token=bool:true --special --override-kv "tokenizer.ggml.template_suffix=str:<think>"