No GPU left behind: Unlocking Efficiency with Co-located vLLM in TRL

Source: HuggingFace Blog · June 3, 2025