Model Examples

Check out the coreai-models project for model examples with Core AI. Several of the LLM models there utilize coreai-opt for compression, via various compression configs (4 bit configs, mixed precision 4/8 bit configs etc).

The examples in this section use simpler models (like ResNet50) to illustrate various compression options, hyper-parameters choices and processes that you may use for determining which methods to apply to your specific model.

• ResNet50 model: Weight + Activation Quantization (PTQ)
• Mixed-precision palettization with ResNet50
• EDSR model: Weight palettization + activation quantization (PTQ)