Fast and Efficient 2-bit LLM Inference on GPU: 2/4/16-bit in a Weight Matrix with Asynchronous Dequantization
This work addresses the challenge of expensive LLM inference for AI practitioners by providing an incremental improvement in quantization techniques.
The paper tackles the problem of high inference cost in large language models (LLMs) by proposing a mixed-precision quantization method with asynchronous dequantization, achieving a 1.74x end-to-end speedup for Llama2-7b and reducing runtime and total costs by up to 2.53x and 2.29x with negligible accuracy loss.
Large language models (LLMs) have demonstrated impressive abilities in various domains while the inference cost is expensive. Many previous studies exploit quantization methods to reduce LLM inference cost by reducing latency and memory consumption. Applying 2-bit single-precision weight quantization brings >3% accuracy loss, so the state-of-the-art methods use mixed-precision methods for LLMs (e.g. Llama2-7b, etc.) to improve the accuracy. However, challenges still exist: (1) Uneven distribution in weight matrix. (2) Large speed degradation by adding sparse outliers. (3) Time-consuming dequantization operations on GPUs. To tackle these challenges and enable fast and efficient LLM inference on GPUs, we propose the following techniques in this paper. (1) Intra-weight mixed-precision quantization. (2) Exclusive 2-bit sparse outlier with minimum speed degradation. (3) Asynchronous dequantization. We conduct extensive experiments on different model families (e.g. Llama3, etc.) and model sizes. We achieve 2.91-bit for each weight considering all scales/zeros for different models with negligible loss. As a result, with our 2/4/16 mixed-precision quantization for each weight matrix and asynchronous dequantization during inference, our design achieves an end-to-end speedup for Llama2-7b is 1.74x over the original model, and we reduce both runtime cost and total cost by up to 2.53x and 2.29x with less GPU requirements.