---
language:
- zh
- en
license: other
license_name: upstream-code-licenses
license_link: https://huggingface.co/datasets/yuyan-lang/codeparrot-github-code-chinese-translation-v1/blob/main/%E8%AE%B8%E5%8F%AF%E4%B8%8E%E6%9D%A5%E6%BA%90.md
size_categories:
- 100M<n<1B
task_categories:
- text-generation
pretty_name: CodeParrot Chinese Dictionary Substitution v1
configs:
- config_name: default
  data_files:
  - split: train
    path: 数据/*.parquet
---

# CodeParrot GitHub Code Chinese Translation v1

**Deterministic dictionary substitution, not LLM translation.** Natural Chinese fluency, semantic equivalence and executable code are not guaranteed.

**Release v1:** All 1,126 shards have been uploaded. Remote shard counts and byte sizes match the local manifest; see 发布状态.json.

[English algorithm](生成算法.英语.md) · [现代汉语](生成算法.汉语.md) · [先秦文言](生成算法.文言.md)

## Load / 使用

```python
from datasets import load_dataset
stream = load_dataset("yuyan-lang/codeparrot-github-code-chinese-translation-v1", streaming=True, split="train")
row = next(iter(stream))
print(row["content_中文"])
```

Original English columns are retained alongside Chinese derivatives. Only the three translated columns aim to eliminate ASCII alphabetic characters; the dataset as a whole contains English.

See [validation](检查报告.json), [shard mapping](分片清单.json), [provenance and licenses](许可与来源.md), and the reproducible source under 程序/. The three algorithm documents describe the final implementation; historical LLM experiments are background only.