Model Optimization

Quantization

• Transforming an ML program into an approximated representation with available lower precision operations
• Convert Float32 to Int8
• Weights, activations