---
library_name: transformers
tags: []
---

expanded 49556 tokens

Following [The Optimal Vocabulary Size Predictor](https://huggingface.co/spaces/sail/scaling-with-vocab-demo), I recommend using this tokenizer with 13B model.