Datasets:
Tasks:
Text Retrieval
Modalities:
Text
Sub-tasks:
document-retrieval
Languages:
code
Size:
100K - 1M
ArXiv:
License:
update repobench
Browse files- README.md +4 -48
- repobench-r.py +1 -1
README.md
CHANGED
|
@@ -62,56 +62,12 @@ The dataset supports two programming languages, Python and Java, and contains tw
|
|
| 62 |
- `cff`: short for `cross_file_first`, where the cross-file module in next line is first used in the current file.
|
| 63 |
- `cfr`: short for `cross_file_random`, where the cross-file module in next line is not first used in the current file.
|
| 64 |
|
| 65 |
-
For each setting, we provide
|
| 66 |
-
|
| 67 |
-
So, the specific information of each split is as follows:
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
<table>
|
| 71 |
-
<tr>
|
| 72 |
-
<th>subset</th>
|
| 73 |
-
<th>split</th>
|
| 74 |
-
<th>#</th>
|
| 75 |
-
</tr>
|
| 76 |
-
<tr>
|
| 77 |
-
<td rowspan="2">python_cff</td>
|
| 78 |
-
<td>easy</td>
|
| 79 |
-
<td>8,000</td>
|
| 80 |
-
</tr>
|
| 81 |
-
<tr>
|
| 82 |
-
<td>hard</td>
|
| 83 |
-
<td>8,000</td>
|
| 84 |
-
</tr>
|
| 85 |
-
<tr>
|
| 86 |
-
<td rowspan="2">python_cfr</td>
|
| 87 |
-
<td>easy</td>
|
| 88 |
-
<td>4,000</td>
|
| 89 |
-
</tr>
|
| 90 |
-
<tr>
|
| 91 |
-
<td>hard</td>
|
| 92 |
-
<td>4,000</td>
|
| 93 |
-
</tr>
|
| 94 |
-
<tr>
|
| 95 |
-
<td rowspan="2">java_cff</td>
|
| 96 |
-
<td>easy</td>
|
| 97 |
-
<td>8,000</td>
|
| 98 |
-
</tr>
|
| 99 |
-
<tr>
|
| 100 |
-
<td>hard</td>
|
| 101 |
-
<td>8,000</td>
|
| 102 |
-
</tr>
|
| 103 |
-
<tr>
|
| 104 |
-
<td rowspan="2">java_cfr</td>
|
| 105 |
-
<td>easy</td>
|
| 106 |
-
<td>4,000</td>
|
| 107 |
-
</tr>
|
| 108 |
-
<tr>
|
| 109 |
-
<td>hard</td>
|
| 110 |
-
<td>4,000</td>
|
| 111 |
-
</tr>
|
| 112 |
-
</table>
|
| 113 |
|
|
|
|
| 114 |
|
|
|
|
|
|
|
| 115 |
|
| 116 |
|
| 117 |
|
|
|
|
| 62 |
- `cff`: short for `cross_file_first`, where the cross-file module in next line is first used in the current file.
|
| 63 |
- `cfr`: short for `cross_file_random`, where the cross-file module in next line is not first used in the current file.
|
| 64 |
|
| 65 |
+
For each setting, we provide `train` and `test` subset, and each subset with two levels of difficulty: `easy` and `hard`.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 66 |
|
| 67 |
+
Suppose the number of code snippets in the context is k,
|
| 68 |
|
| 69 |
+
- For the `easy` subset, we have \\(5 \leq k < 10\\)
|
| 70 |
+
- For the `hard` subset, we have \\(k \geq 10\\).
|
| 71 |
|
| 72 |
|
| 73 |
|
repobench-r.py
CHANGED
|
@@ -149,7 +149,7 @@ class RepoBenchR(datasets.GeneratorBasedBuilder):
|
|
| 149 |
"import_statement": example["import_statement"],
|
| 150 |
"code": example["code"],
|
| 151 |
"next_line": example["next_line"],
|
| 152 |
-
"gold_snippet_index": example["
|
| 153 |
}
|
| 154 |
|
| 155 |
|
|
|
|
| 149 |
"import_statement": example["import_statement"],
|
| 150 |
"code": example["code"],
|
| 151 |
"next_line": example["next_line"],
|
| 152 |
+
"gold_snippet_index": example["golden_snippet_index"]
|
| 153 |
}
|
| 154 |
|
| 155 |
|