XLA
섬네일|openxla XLA[1] (Accelerated Linear Algebra) is an open-source machine learning (ML) compiler for GPUs, CPUs, and ML accelerators.
The XLA compiler takes models from popular ML frameworks such as PyTorch, TensorFlow, and JAX, and optimizes them for high-performance execution across different hardware platforms including GPUs, CPUs, and ML accelerators.
XLA for TensorFlow
In many cases, XLA improves performance over native TensorFlow. The major difference between these two is the fusion optimizer in XLA. Instead of executing many small kernels back to back, XLA optimizes these into larger kernels. This greatly reduces execution time of bandwidth bound kernels. XLA also offers many algebraic simplifications, far superior to what Tensorflow offers.
There are comprehensive instruction how to utilize XLA for TensorFlow