Failed to set the param enable-thinking = True, error: 'xllamacpp.xllamacpp.CommonParams' object has no attribute 'enable-thinking' #3860
Description
我使用卡片点击的配置方式，将各类配置以及参数填入后，点击小火箭图标可以正常启动，且没有报错；
然后我将这些配置复制为命令行参数之后，进入到xinference镜像构建的容器中，执行launch命令之后，就会报错，报参数无法传递进去的错误。
这是我在容器内执行的命令：
xinference launch --model-name qwen3 --model-type LLM --model-engine llama.cpp --model-format ggufv2 --size-in-billions 1_7 --quantization Q8_0 --replica 1 --n-worker 1 --model-path models/Qwen3-1.7B-GGUF/Qwen3-1.7B-Q8_0.gguf --enable-thinking true --reasoning-content false --n_parallel 64
以下是报错：
2025-07-29 16:00:13,048 xinference.core.model 2745 INFO Start requests handler.
2025-07-29 16:00:13,072 xinference.model.llm.llama_cpp.core 2745 ERROR Failed to set the param trust_remote_code = True, error: 'xllamacpp.xllamacpp.CommonParams' object has no attribute 'trust_remote_code'
2025-07-29 16:00:13,072 xinference.model.llm.llama_cpp.core 2745 ERROR Failed to set the param enable-thinking = True, error: 'xllamacpp.xllamacpp.CommonParams' object has no attribute 'enable-thinking'
2025-07-29 16:00:13,072 xinference.model.llm.llama_cpp.core 2745 ERROR Failed to set the param reasoning-content = False, error: 'xllamacpp.xllamacpp.CommonParams' object has no attribute 'reasoning-content'
请问这些参数应该怎么传递进去呢，比如enable-thinking、Presence_penalty (在llamacpp中好像是叫repeat-penalty)；因为qwen3-1.7b-gguf模型对话时容易陷入死循环，考虑传递这些参数，但是发现传递不进去。