Skip to content

LLM Configuration

In some cases, it is helpful to adjust LLM sampling parameters (e.g., temperature, top-p, top-k, maximum new tokens) or use reasoning models (e.g., OpenAI o-series models, Qwen3) which requires special treatments in system prompt, user prompt, and sampling parameters. For example, OpenAI o-series reasoning models disallow passing a system prompt or setting custom temperature. Another example is Qwen3 hybrid thinking mode. Special tokens "/think" and "/no_think" should be appended to user prompts to control for the reasoning behavior.

Config class LLMs
BasicLLMConfig Most non-reasoning LLMs
  • Llama4
  • Qwen3-30B-A3B-Instruct-2507
  • ...
    		•
    ReasoningLLMConfig Most reasoning LLMs
  • Qwen3-30B-A3B-Thinking-2507
  • gpt-oss-120b
  • ...
    		•
    Qwen3LLMConfig Qwen3 hybrid thinking
  • Qwen3-30B-A3B
  • Qwen3-32B
  • ...
    		•
    OpenAIReasoningLLMConfig OpenAI API reasoning models
  • "o" series (o1, o3, o4)
  • ...
    		•

    Setting sampling parameters

    LLM sampling parameters such as temperature, top-p, top-k, and maximum new tokens can be set by passing a LLMConfig class to the InferenceEngine constructor.

    from llm_ie.engines import OpenAIInferenceEngine, BasicLLMConfig

config = BasicLLMConfig(temperature=0.2, max_new_tokens=4096)
inference_engine = OpenAIInferenceEngine(model="gpt-4.1-mini", config=config)

    Reasoning models

    To use reasoning models such as OpenAI o-series (e.g., o1, o3, o3-mini, o4-mini), some special processing is required. We provide dedicated configuration classes for them.

    General reasoning models

    Most reasoning models can be configured with ReasoningLLMConfig. By specifying the start and end thinking tags, the reasoning tokens will be excluded from model response while stored in messages log (if return_messages_log=True).

    from llm_ie.engines import VLLMInferenceEngine, ReasoningLLMConfig

llm = VLLMInferenceEngine(model="Qwen/Qwen3-30B-A3B-Thinking-2507", 
                          config=ReasoningLLMConfig(thinking_token_start="<think>", thinking_token_end="</think>", temperature=0.8, max_new_tokens=8192))

    OpenAI o-series reasoning models

    OpenAI o-series reasoning model API does not allow setting system prompts. Contents in the system should be included in user prompts. Also, custom temperature is not allowed. We provide a dedicated configuration class OpenAIReasoningLLMConfig for these models. 

    from llm_ie.engines import OpenAIInferenceEngine, OpenAIReasoningLLMConfig

inference_engine = OpenAIInferenceEngine(model="o4-mini", 
                                         config=OpenAIReasoningLLMConfig(reasoning_effort="low"))

    Qwen3 (hybrid thinking mode)

    Does NOT work for Qwen3 2507 models. Use ReasoningLLMConfig and BasicLLMConfig instead.

    Qwen3 has a special way to manage reasoning behavior. The same models have thinking mode and non-thinking mode, controled by the prompting template. When a special token "/think" is appended to the user prompt, the models generate thinking tokens in a <think>... </think> block. When a special token "/no_think" is appended to the user prompt, the models generate an empty <think>... </think> block. We provide a dedicated configuration class Qwen3LLMConfig for these models. 

    from llm_ie.engines import VLLMInferenceEngine, Qwen3LLMConfig

# Thinking mode
llm = VLLMInferenceEngine(model="Qwen/Qwen3-30B-A3B", 
                          config=Qwen3LLMConfig(thinking_mode=True, temperature=0.6, top_p=0.95, top_k=20, max_new_tokens=8192))

# Non-thinking mode
llm = VLLMInferenceEngine(model="Qwen/Qwen3-30B-A3B", 
                          config=Qwen3LLMConfig(thinking_mode=False, temperature=0.0, max_new_tokens=2048))