NIRVLab — WordPiece Tokenizer for Swahili XNLI
A WordPiece tokenizer (BERT-style) trained from scratch on the Swahili (sw) subset
of the facebook/xnli dataset.
Training Details
| Parameter | Value |
|---|---|
| Algorithm | WordPiece |
| Vocabulary size | 32,000 |
| Special tokens | <s>, <pad>, </s>, <unk>, <mask> |
| Corpus | facebook/xnli / sw — all splits |
| Corpus size | 800,404 sentences |
| Normalizer | NFC |
| Pre-tokenizer | Whitespace |
| Min frequency | 2 |
| Continuing subword prefix | ## |
Evaluation Metrics
| Metric | Value |
|---|---|
| Tokens / char | 0.2105 |
| Fertility (tokens / word) | 1.3605 |
| Avg sequence length | 16.58 tokens |
| Vocabulary coverage | 1.0000 |
Usage
from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("NIRVLab/xnli-wordpiece-sw-32k")
tokens = tokenizer("Habari ya leo!", return_tensors="pt")
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support