Understanding the Impact of Post-Training Quantization on Large Language Models
Large language models (LLMs) are rapidly increasing in size, with the number of parameters becoming a key factor in the success of many commercial models, such as ChatGPT, Claude, and Bard. Even the recently released publicly accessible models for commercial usage, such as Falcon and Llama2, come equipped with billions of parameters. This significant increase in the number of parameters makes deployment and operation very costly. The remarkable progress in the field of quantization for large neural networks in general and LLMs in particular, has made these models more accessible by enabling them to be deployed on consumer-grade GPUs. Quantized models generally demonstrate comparable performance levels to their unquantized base counterparts. Nonetheless, there exists a notable gap in our comprehensive understanding of how these quantized models respond to hyperparameters, such as temperature, max new tokens, and topk, particularly for next word prediction. The present analysis reveals that nf4 and fp4 are equally proficient 4-bit quantization techniques, characterized by similar attributes such as inference speed, memory consumption, and the quality of generated content. Nevertheless, these quantization methods exhibit distinct behaviors at varying temperature settings, both in the context of smaller and larger models. It is noteworthy that, in general, 4-bit quantized models of varying sizes exhibit heightened sensitivity to lower temperature settings, unlike their unquantized counterparts. Additionally, int8 quantization is associated with significantly slower inference speeds, whereas unquantized fp16 models consistently yield the fastest inference speeds across models of all sizes.
READ FULL TEXT