---
license: mit
language:
- arb  # ISO 639-3 code or "und" if not identifiable
tags:
- tokenizer
- bpe
- flexitok
- fineweb2
---

# Byte-Level BPE Tokenizer: arb_Arab (4K)

A **Byte-Level BPE** tokenizer trained on **arb_Arab** data from Fineweb-2-HQ.

## Training Details

| Parameter | Value |
|-----------|-------|
| Algorithm | Byte-Level BPE |
| Language | `arb_Arab` |
| Target Vocab Size | 4,000 |
| Final Vocab Size | 4,000 |
| Pre-tokenizer | custom:arb_Arab |
| Number handling | individual |
| Contraction handling | True |
| Normalizer | NFC |
| Special Tokens | `<s>`, `</s>`, `<pad>`, `<unk>` |
| Training Shards | 2 |

## Usage

```python
from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("flexitok/bpe_arb_Arab_4000_v2")
tokens = tokenizer.encode("Hello, world!")
```

## Files

- `tokenizer.json` — Full HuggingFace tokenizer
- `vocab.json` — Vocabulary mapping
- `merges.txt` — BPE merge rules

## Sample Encoding
| Text | Tokens | Token IDs |
|------|--------|-----------|
| `Hello, world! 12345 This is a test. こんにちは` | `H, ell, o, ,, Ġ, w, or, l, d, !, Ġ, 12, 3, 45, Ġ, T, h, is, Ġ, is` | `43, 3814, 82, 15, 178, 90, 691, 79, 71, 4, 178, 1214, 22, 3735, 178, 55, 75, 855, 178, 855` |