Helion is a PyTorch-native domain-specific language (DSL) for authoring performance portable machine learning kernels.
Helion is a PyTorch-native domain-specific language (DSL) for authoring high-performance machine learning kernels. It dramatically simplifies the development of fast, scalable ML kernels by bridging the gap between high-level productivity and low-level control—offering the user-friendly simplicity of PyTorch with the performance of lower-level languages. Helion’s core innovations include PyTorch-native syntax that lets developers write kernels using familiar idioms, advanced autotuning that automatically explores a vast configuration space to find optimal implementations, and hardware portability that enables a single kernel definition to run efficiently across GPUs and hardware accelerators.