Skip to content

Latest commit

 

History

History
30 lines (20 loc) · 575 Bytes

README.md

File metadata and controls

30 lines (20 loc) · 575 Bytes

Mini bpe in Rust

Rust

Port minbpe to rust as learning process

Benchmark

Build binary

cargo build --release

Run tokenizer

./target/release/rbpe --tokenizer {basic, regex}

Results

On my m1 book, I got:

Mode Time (rust) Time (python)
Basic 0.4s 5.65s
Regex 1.23s 9.01s