Palettization

This section describes the following:

• Palettization Overview: An overview of palettization (aka weight clustering) and considerations on model size and performance.
• Post-Training Palettization: Weight clustering compression of Core ML models using ct.optimize.coreml.palettize_weights.
• Training-Time Palettization : Weight clustering compression on PyTorch models with data and fine-tuning, using ct.optimize.torch.palettization.DKMPalettizer.