How does pruning/ quantization affect model accuracy, and how can I debug accuracy drops on silicon?

Quantization (INT8, W4A8, mixed precision, etc.) and unstructured sparsity reduce memory and latency but can introduce negligible accuracy losses depending on the model and data distribution. The models on model garden are curated models from Ambarella with tradeoff between data formats and pruning budget to recover accuracy. Accuracy of the models can be verified on boards.