GPUTOK: GPU Accelerated Byte Level BPE Tokenization

TL;DR

This paper introduces GPUTOK, a GPU-accelerated byte-level BPE tokenizer that significantly speeds up tokenization for large language models while maintaining output quality, enabling more efficient long-context processing.

Contribution

The authors developed a GPU-based BPE tokenizer with optimized kernel implementations that outperform existing CPU tokenizers in speed, suitable for large-scale language model applications.

Findings

GPU tokenizer is 1.7x faster than tiktoken on WikiText103 sequences

It is about 7.6x faster than HuggingFace GPT-2 tokenizer on long inputs

Output quality remains comparable to existing tokenizers within 1% similarity

Abstract

As large language models move toward million-token context windows, CPU tokenizers become a major slowdown because they process text one step at a time while powerful GPUs sit unused. We built a GPU-based byte-level BPE tokenizer that follows GPT-2's merge rules. It includes a basic BlockBPE-style kernel and a faster, optimized version that uses cuCollections static map, CUB reductions, and a pybind11 interface for Python. On WikiText103 sequences up to 131k tokens, the optimized GPU tokenizer produces the same tokens as a CPU version and, for the longest inputs, is about 1.7x faster than tiktoken and about 7.6x faster than the HuggingFace GPT-2 tokenizer. Nsight profiling shows that 70-80% of CUDA API time goes to memory allocation, so adding memory pooling should give the biggest speed boost next. Tests on generation tasks using WikiText103 prompts show that our GPU tokenizer's outputs stay within about one percentage point of tiktoken and HuggingFace GPT-2 on similarity and overlap metrics, meaning it keeps output quality while making long-context inference more practical.