Skip to main content

Compiler

Tools and APIs for compiling custom ONNX models to run on Metis hardware. Compilation converts a floating-point model to an int8 binary optimized for the AIPU.

PageWhat it covers
Compiler CLIThe compile command-line tool. Covers basic usage, quantize-only mode, real-image calibration, output artifacts, and error status codes.
Compiler Python APIThe compiler.quantize() and compiler.compile() Python API. Covers CompilerConfig, the two-step quantize → compile workflow, and usage examples.
Compiler ConfigurationMulti-core compilation modes: Batch-1 (independent cores, lower latency) vs Batch-4 (shared memory, higher throughput). Includes resource allocation for multi-model pipelines.
ONNX Operator SupportWhich ONNX operators are accelerated on the AIPU (opsets 14–17), which are constrained, and what falls back to CPU.
CompilerConfig ReferenceFull property listing for CompilerConfig: all quantization, scheduling, memory, and hardware parameters with types, defaults, and enum values.

See the Deploy Custom Weights tutorial for the end-to-end workflow.