TinyML

Quantization

Technique for reducing the precision of neural network weights and activations from 32 bits to 8 bits or less, significantly reducing memory and computational requirements.

← Geri