Hugging Face Blog·· 2025-09-02AI score36
How to Speed Up ZeroGPU Spaces by 1.3–1.8× Using PyTorch Ahead-of-Time Compilation
Make your ZeroGPU Spaces go brrr with ahead-of-time compilation
AI briefing
Hugging Face introduces using PyTorch ahead-of-time (AoT) compilation instead of torch.compile in ZeroGPU Spaces to avoid repeated compilation in ZeroGPU's short-lived processes, achieving 1.3×–1.8× speedups for models such as Flux, Wan, and LTX.
Open original huggingface.co