Quantization

8-bit Quantization

Compression technique reducing model weights from 32 bits to 8 bits, offering an optimal trade-off between performance and accuracy for LLMs.

← Indietro