The steps to run model inference in gpt-oss-recipes are as follows:
- Loading Models and Splitters: Use
AutoModelForCausalLMcap (a poem)AutoTokenizerLoad the specified model (e.g.openai/gpt-oss-20b). - input prompt: Define user messages (e.g.
"如何用Python写一个排序算法？"), and throughapply_chat_templatemethod to process the input. - Generate results: Call
model.generatemethod generates the response and decodes the output using a splitter. - Adjustment of inference parameters (optional): The level of reasoning detail can be adjusted by system prompts, such as setting the
"Reasoning: high"to generate a more detailed reasoning process.
The reasoning example scripts are usually located in theinference.pyfile, the model returns the generated results after running.
This answer comes from the articleCollection of scripts and tutorials for fine-tuning OpenAI GPT OSS modelsThe
May not be reproduced without permission:AI productivity tools " How to run model inference in gpt-oss-recipes?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil