API Overview

This section provides an overview of the coremltools.optimizeAPI and the updates to coremltools.convert to support the conversion of models with weight compression:

• For data-free model optimization of Core ML models, see optimize.coreml API overview.
• For data-dependent optimizations that require fine-tuning the PyTorch model, see the optimize.torch API overview.
• For conversion options for models that are already compressed or quantized, either using coremltools.optimize.torch or otherwise (using source framework's native APIs), see Converting compressed source models.

If you want to get familiar with the different compression techniques, see the following overviews first:

• Pruning Overview
• Palettization Overview
• Quantization Overview