[2101.01321] I-BERT: Integer-only BERT Quantization

Abstract:Transformer based models, like BERT and RoBERTa, have achieved state-of-the-art results in many Natural Language Processing tasks. However, their memory footprint, inference latency, and power consumption are prohibitive efficient inference at the edge, and even at the data center. While quantization can be a viable solution for this, previous work on quantizing Transformer based models use floating-point arithmetic during inference, which cannot efficiently utilize integer-only logical units such as the recent Turing Tensor Cores, or traditional integer-only ARM processors. In this work, we propose I-BERT, a novel quantization scheme for Transformer based models that quantizes the entire inference with integer-only arithmetic. Based on lightweight integer-only approximation methods for nonlinear operations, e.g., GELU, Softmax, and Layer Normalization, I-BERT performs an end-to-end integer-only BERT inference without any floating point calculation. We evaluate our approach on GLUE downstream tasks using RoBERTa-Base/Large. We show that for both cases, I-BERT achieves similar (and slightly higher) accuracy as compared to the full-precision baseline. Furthermore, our preliminary implementation of I-BERT shows a speedup of 2.4-4.0x for INT8 inference on a T4 GPU system as compared to FP32 inference. The framework has been developed in PyTorch and has been open-sourced.

Abstract:Transformer based models, like BERT and RoBERTa, have achieved state-of-the-art results in many Natural Language Processing tasks. However, their memory footprint, inference latency, and power consumption are prohibitive efficient inference at the edge, and even at the data center. While quantization can be a viable solution for this, previous work on quantizing Transformer based models use floating-point arithmetic during inference, which cannot efficiently utilize integer-only logical units such as the recent Turing Tensor Cores, or traditional integer-only ARM processors. In this w

related reading

• LLM.int8()arxiv.org
• Transformer Inference Arithmetic | kipply's blogkipp.ly
• A Mathematical Framework for Transformer Circuitstransformer-circuits.pub
• Accelerated Inference for Large Transformer Models Using NVIDIA Triton Inference Server | NVIDIA Technical Blogdeveloper.nvidia.com
• [2207.00032] DeepSpeed Inference: Enabling Efficient Inference of Transformer Models at Unprecedented Scalearxiv.org
• Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluationarxiv.org
• Transformers Inference Optimization Toolset | AstraBlogastralord.github.io
• Achieving FP32 Accuracy for INT8 Inference Using Quantization Aware Training with NVIDIA TensorRT | NVIDIA Technical Blogdeveloper.nvidia.com
• [2011.14203] EdgeBERT: Sentence-Level Energy Optimizations for Latency-Aware Multi-Task NLP Inferencearxiv.org
• The Bitter Lesson is coming for Tokenization – ⛰️ lucalplucalp.dev
• Transformer inference tricks - by Finbarr Timbersartfintel.com
• Transformer Math 101 | EleutherAI Blogblog.eleuther.ai