All terms
Frameworks & Tools
OpenXLA
An open compiler stack that optimizes machine learning models for varied hardware.
Definition
OpenXLA is an open-source compiler stack for machine learning. It takes models expressed in frameworks like JAX, PyTorch, or TensorFlow and compiles them into optimized code for a range of hardware, including GPUs and TPUs. By sitting between frameworks and accelerators, it aims to give portable, high-performance execution without framework-specific tuning.