The repository offers two models:gpt-oss-120b(117B parameters) andgpt-oss-20b(21B parameters). Hardware conditions need to be considered when selecting a model:
- High-performance GPUs (e.g. H100): Recommended use
gpt-oss-120b, requires at least 80GB of GPU memory. - Consumer-grade hardware (e.g., 16GB RAM devices): Recommended use
gpt-oss-20bThe resource footprint is low.
This is done in the script by modifying themodel_pathVariable selection models, for example:
model_path = "openai/gpt-oss-20b" # 选择20B模型
# model_path = "openai/gpt-oss-120b" # 选择120B模型The script automatically configures device mapping and optimization settings based on model size.
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 choose the right model in the gpt-oss-recipes repository?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil