gpt-oss-20b | NVIDIA NeMo AutoModel

GPT-OSS is OpenAI’s open-weight model family featuring QuickGELU activations and activation clamping for training stability.

The documented workflows cover full-parameter fine-tuning, LoRA, packed sequences, and a DGX Spark LoRA configuration.

Set up NeMo AutoModel with the latest container or follow the installation instructions.

Fine-Tune gpt-oss-20b

From the repository root, run:

uv run automodel --nproc-per-node=8 examples/llm_finetune/gpt_oss/gpt_oss_20b.yaml

Goal	Start Here
Full-parameter fine-tuning	Use the full-parameter recipe.
LoRA fine-tuning	Use the LoRA recipe.
Fine-tuning with 1,024-token packed sequences	Use the packed-sequence recipe.
LoRA fine-tuning on DGX Spark	Use the DGX Spark recipe.

Model	HF ID
GPT-OSS 20B	`openai/gpt-oss-20b`