What AI says about SFTTrainer

Hugging Face's TRL library provides a Supervised Fine-Tuning (SFT) Trainer for fine-tuning language models. The SFTTrainer supports standard language modeling, prompt-completion, and conversational datasets, and will automatically apply a chat template when given conversational data. The quick-start example shows training a model like Qwen/Qwen3-0.6B on the Capybara dataset, including guidance on preprocessing datasets into the required formats.