File size: 1,233 Bytes
9765bfb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
# NLLB Base CoreML Model

NLLB-200 (No Language Left Behind) の distilled 600M モデルを CoreML 形式に変換したものです。

## モデル情報

- **ベースモデル**: facebook/nllb-200-distilled-600M
- **対応言語**: 200言語(eng_Latn ↔ jpn_Jpan 等)
- **フォーマット**: CoreML (.mlpackage)
- **精度**: Float16
- **最小デプロイメントターゲット**: macOS 13.0

## ファイル構成

```
coreml/
├── NLLBEncoder.mlpackage/    # Encoder モデル
├── NLLBDecoder.mlpackage/    # Decoder モデル
├── sentencepiece.bpe.model   # トークナイザー
└── tokenizer_config.json     # トークナイザー設定
```

## 使用方法

```swift
import CoreML

// モデル読み込み
let encoder = try MLModel(contentsOf: encoderURL)
let decoder = try MLModel(contentsOf: decoderURL)

// 推論
// ... (TranslationEngine.swift を参照)
```

## 言語コード

- `eng_Latn`: 256047 (英語)
- `jpn_Jpan`: 256079 (日本語)

## ライセンス

CC-BY-NC-4.0 (元モデルのライセンスに従う)

## 関連リンク

- [NLLB-200](https://huggingface.co/facebook/nllb-200-distilled-600M)
- [translation_subtitle](https://github.com/nawta/translation_subtitle)