Initialize Plaincode CNL dataset card
Browse files
README.md
ADDED
|
@@ -0,0 +1,158 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
pretty_name: Plaincode CNL 100k
|
| 3 |
+
language:
|
| 4 |
+
- code
|
| 5 |
+
- en
|
| 6 |
+
- es
|
| 7 |
+
- fr
|
| 8 |
+
- pt
|
| 9 |
+
- zh
|
| 10 |
+
- hi
|
| 11 |
+
tags:
|
| 12 |
+
- code
|
| 13 |
+
- python
|
| 14 |
+
- plaincode
|
| 15 |
+
- controlled-natural-language
|
| 16 |
+
- semantic-roundtrip
|
| 17 |
+
size_categories:
|
| 18 |
+
- 100K<n<1M
|
| 19 |
+
configs:
|
| 20 |
+
- config_name: default
|
| 21 |
+
data_files:
|
| 22 |
+
- split: train
|
| 23 |
+
path: data/train-*.jsonl.gz
|
| 24 |
+
---
|
| 25 |
+
|
| 26 |
+
# Plaincode CNL 100k
|
| 27 |
+
|
| 28 |
+
This dataset is generated with Plaincode by streaming Python source files, projecting each accepted Python module/function into the current controlled-natural-language surfaces, and enforcing strict exact Python roundtrip proof for every emitted language.
|
| 29 |
+
|
| 30 |
+
## Included surfaces
|
| 31 |
+
|
| 32 |
+
- Python source
|
| 33 |
+
- English CNL (`en`)
|
| 34 |
+
- Spanish CNL (`es`)
|
| 35 |
+
- French CNL (`fr`)
|
| 36 |
+
- Portuguese CNL (`pt`)
|
| 37 |
+
- Mandarin CNL (`zh`)
|
| 38 |
+
- Hindi CNL (`hi`)
|
| 39 |
+
|
| 40 |
+
## Acceptance gate
|
| 41 |
+
|
| 42 |
+
A row is accepted only when every requested CNL surface reverses back to the exact original Python bytes through Plaincode proof:
|
| 43 |
+
|
| 44 |
+
- `sha_exact_ok == true`
|
| 45 |
+
- `char_exact_ok == true`
|
| 46 |
+
- `byte_exact_ok == true`
|
| 47 |
+
- `exact_ok == true`
|
| 48 |
+
- `reverse_ast_ok == true`
|
| 49 |
+
- `reverse_compile_ok == true`
|
| 50 |
+
|
| 51 |
+
The public CNL fields must not contain private CST/proof transport markers.
|
| 52 |
+
|
| 53 |
+
## Schema
|
| 54 |
+
|
| 55 |
+
Each row contains:
|
| 56 |
+
|
| 57 |
+
- `row_id`
|
| 58 |
+
- `source`: Python code, sha256, counts, compile status, and source metadata
|
| 59 |
+
- `plaincode_cnl`: `{lang: cnl_text}` for `en/es/fr/pt/zh/hi`
|
| 60 |
+
- `roundtrip_python`: `{lang: recovered_python}`
|
| 61 |
+
- `proof`: strict proof dictionary per language
|
| 62 |
+
- `strict_exact_by_language`
|
| 63 |
+
- `semantic_family`: AST-derived tags for diversity inspection
|
| 64 |
+
- `row_sha256`
|
| 65 |
+
|
| 66 |
+
## Generation stats
|
| 67 |
+
|
| 68 |
+
```json
|
| 69 |
+
{
|
| 70 |
+
"accepted_rows": 0,
|
| 71 |
+
"completed_shards": [],
|
| 72 |
+
"duplicates": 0,
|
| 73 |
+
"family_counts": {},
|
| 74 |
+
"filter_reasons": {},
|
| 75 |
+
"filtered_candidates": 0,
|
| 76 |
+
"hf_repo_id": "CircularBalls/plaincode-cnl-100k",
|
| 77 |
+
"languages": [
|
| 78 |
+
"en",
|
| 79 |
+
"es",
|
| 80 |
+
"fr",
|
| 81 |
+
"pt",
|
| 82 |
+
"zh",
|
| 83 |
+
"hi"
|
| 84 |
+
],
|
| 85 |
+
"projection_failures": 0,
|
| 86 |
+
"run_id": "run_20260528_214053_utc",
|
| 87 |
+
"seen_candidates": 0,
|
| 88 |
+
"shards_completed": 0,
|
| 89 |
+
"source_label_counts": {},
|
| 90 |
+
"source_specs": [
|
| 91 |
+
{
|
| 92 |
+
"config": "Python-mit",
|
| 93 |
+
"dataset": "codeparrot/github-code-clean",
|
| 94 |
+
"enabled": true,
|
| 95 |
+
"label": "github_code_clean_python_mit",
|
| 96 |
+
"split": "train",
|
| 97 |
+
"trust_remote_code": false,
|
| 98 |
+
"weight": 4
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"config": "Python-apache-2.0",
|
| 102 |
+
"dataset": "codeparrot/github-code-clean",
|
| 103 |
+
"enabled": true,
|
| 104 |
+
"label": "github_code_clean_python_apache_2_0",
|
| 105 |
+
"split": "train",
|
| 106 |
+
"trust_remote_code": false,
|
| 107 |
+
"weight": 3
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"config": "Python-bsd-3-clause",
|
| 111 |
+
"dataset": "codeparrot/github-code-clean",
|
| 112 |
+
"enabled": true,
|
| 113 |
+
"label": "github_code_clean_python_bsd_3",
|
| 114 |
+
"split": "train",
|
| 115 |
+
"trust_remote_code": false,
|
| 116 |
+
"weight": 2
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"config": "Python-bsd-2-clause",
|
| 120 |
+
"dataset": "codeparrot/github-code-clean",
|
| 121 |
+
"enabled": true,
|
| 122 |
+
"label": "github_code_clean_python_bsd_2",
|
| 123 |
+
"split": "train",
|
| 124 |
+
"trust_remote_code": false,
|
| 125 |
+
"weight": 2
|
| 126 |
+
},
|
| 127 |
+
{
|
| 128 |
+
"config": "Python-isc",
|
| 129 |
+
"dataset": "codeparrot/github-code-clean",
|
| 130 |
+
"enabled": true,
|
| 131 |
+
"label": "github_code_clean_python_isc",
|
| 132 |
+
"split": "train",
|
| 133 |
+
"trust_remote_code": false,
|
| 134 |
+
"weight": 1
|
| 135 |
+
},
|
| 136 |
+
{
|
| 137 |
+
"config": "Python-cc0-1.0",
|
| 138 |
+
"dataset": "codeparrot/github-code-clean",
|
| 139 |
+
"enabled": true,
|
| 140 |
+
"label": "github_code_clean_python_cc0",
|
| 141 |
+
"split": "train",
|
| 142 |
+
"trust_remote_code": false,
|
| 143 |
+
"weight": 1
|
| 144 |
+
}
|
| 145 |
+
],
|
| 146 |
+
"started_utc": "2026-05-28T21:40:55+00:00",
|
| 147 |
+
"strict_gate_failures": 0,
|
| 148 |
+
"target_accepted_rows": 100000
|
| 149 |
+
}
|
| 150 |
+
```
|
| 151 |
+
|
| 152 |
+
## Intended use
|
| 153 |
+
|
| 154 |
+
Training/evaluating semantic projection, code-to-CNL, CNL-to-code, multilingual controlled-language alignment, and exact roundtrip experiments.
|
| 155 |
+
|
| 156 |
+
## Source/license note
|
| 157 |
+
|
| 158 |
+
Rows retain source metadata where available, including license fields exposed by upstream datasets. The default generator configuration prioritizes permissive-ish Python source configs. Review upstream licenses before redistribution or commercial use.
|