AI Glossary

ML Compiler

Software that optimizes machine learning model computations for specific hardware, translating high-level model descriptions into efficient low-level instructions.

Key Compilers

XLA: Google's compiler for TensorFlow/JAX. TorchInductor: PyTorch's compiler (torch.compile). TVM: Open-source deep learning compiler. Triton: NVIDIA's GPU programming language for custom kernels.

Why It Matters

ML compilers can improve model performance 2-5x by fusing operations, optimizing memory access patterns, and generating hardware-specific code. torch.compile in PyTorch 2.0 made compilation mainstream.

← Back to AI Glossary

Last updated: March 5, 2026