classical-ja-pretrain / code /scripts /hf_cards /classical-ja-tokenizer.md
kuramot6f's picture
Initial release (history squashed)
7a088d4
|
Raw History Blame Contribute Delete
3.34 kB
---
license: apache-2.0
language:
- ja
- lzh
tags:
- tokenizer
- sentencepiece
- classical-japanese
- literary-chinese
library_name: transformers
---
# Classical Japanese Tokenizer (32k SentencePiece)
A SentencePiece **unigram** tokenizer with 32,000 pieces, trained **only on texts written before 1868**:
classical Japanese, early-modern Japanese and Literary Chinese (漢文). Tokenizers built for modern Japanese
split classical text poorly. This one is fitted to historical kana usage, kanbun and the rare characters
common in premodern books. Byte fallback means it never produces `<unk>`.
It is the tokenizer of [classical-smollm2-trunk](https://huggingface.co/kuramot6f/classical-smollm2-trunk) and
[classical-smollm2-sft](https://huggingface.co/kuramot6f/classical-smollm2-sft).
## Usage
With 🤗 Transformers (the small custom wrapper keeps the exact ID mapping):
```python
from transformers import AutoTokenizer
tok = AutoTokenizer.from_pretrained("kuramot6f/classical-ja-tokenizer", trust_remote_code=True)
enc = tok("春はあけぼの。やうやう白くなりゆく山ぎは")
print(enc["input_ids"]) # starts with BOS (=1)
print(tok.decode(enc["input_ids"], skip_special_tokens=True))
```
With plain SentencePiece:
```python
import sentencepiece as spm
from huggingface_hub import hf_hub_download
sp = spm.SentencePieceProcessor(model_file=hf_hub_download("kuramot6f/classical-ja-tokenizer", "tokenizer.model"))
print(sp.encode("子曰學而時習之", out_type=str))
```
The wrapper differs from plain SentencePiece in one way: a literal `▁` (U+2581) in the input is encoded as
UTF-8 bytes, so decoding always gives back the original text exactly.
## Specs
- Vocabulary 32,000: `<unk>`=0, `<s>`=1, `</s>`=2, `<pad>`=3, plus 256 byte-fallback pieces.
- Unigram model, identity normalization (no NFKC), `character_coverage=0.9995`, `byte_fallback=true`,
SentencePiece 0.2.1.
- Trained on 2M line fragments (70M characters) sampled from 121k documents and 1.33B characters of the
pre-1868 corpus. This is a slightly larger pre-filter version of
[classical-ja-pretrain](https://huggingface.co/kuramot6f/classical-ja-pretrain).
| source | characters per token | byte-token rate | exact round-trip |
|---|---:|---:|---:|
| Kanripo (Chinese classics) | 1.24 | 6.8% | 128/128 |
| Minna de Honkoku (Edo manuscripts) | 1.44 | 11.2% | 128/128 |
| NDL (Edo printed books, OCR) | 1.37 | 0.4% | 128/128 |
| Wikisource (classical literature) | 1.41 | 2.5% | 128/128 |
These are diagnostic samples drawn from the training distribution, not a held-out benchmark.
## Files
`tokenizer.model` / `tokenizer.vocab` (SentencePiece), `classical_tokenizer.py` + `tokenizer_config.json`
(Transformers wrapper), `config.json` / `sampling.json` (training settings), `REPORT.md` (training report,
Japanese), `evaluation.jsonl` / `examples.json` (diagnostics), `manifest.json` (hashes), `train.log`.
## License
Code and tokenizer model: Apache-2.0. The tokenizer was trained on texts under several licenses: public
domain, CC BY-SA 4.0, and CC BY-NC-SA 3.0 TW for CBETA Buddhist texts. See
[classical-ja-pretrain/LICENSES](https://huggingface.co/datasets/kuramot6f/classical-ja-pretrain/blob/main/LICENSES/README.md).
The short text excerpts in `examples.json` / `evaluation.jsonl` keep their source licenses.