Update README.md
Browse files
README.md
CHANGED
@@ -4,19 +4,19 @@ configs:
|
|
4 |
data_files:
|
5 |
- split: test
|
6 |
path: data/python/*.json
|
7 |
-
license: apache-2.0
|
8 |
---
|
9 |
|
10 |
-
# ๐๏ธ Long Code Arena (CI
|
11 |
|
12 |
-
|
|
|
|
|
|
|
13 |
> repository contents in order to make the workflow pass.
|
14 |
|
15 |
-
|
16 |
-
๐๏ธ [**Long Code Arena** benchmark](https://huggingface.co/spaces/JetBrains-Research/long-code-arena).
|
17 |
|
18 |
-
To score your model on this dataset, you can use **CI
|
19 |
-
(https://github.com/JetBrains-Research/lca-baselines/tree/main/ci-fixing/ci-fixing-benchmark)
|
20 |
|
21 |
## How-to
|
22 |
|
@@ -36,7 +36,7 @@ To score your model on this dataset, you can use **CI Fixing benchmark**
|
|
36 |
```
|
37 |
|
38 |
Note that all the data we have is considered to be in the test split.
|
39 |
-
**NOTE**: If you encounter any errors with loading the dataset on Windows, update the datasets library (was tested on datasets==2.16.1)
|
40 |
|
41 |
|
42 |
## Dataset Structure
|
@@ -44,9 +44,9 @@ To score your model on this dataset, you can use **CI Fixing benchmark**
|
|
44 |
This dataset contains logs of the failed GitHub Action workflows for some commits
|
45 |
followed by the commit that passes the workflow successfully.
|
46 |
|
47 |
-
Note that, unlike
|
48 |
|
49 |
-
* Our [CI
|
50 |
their model to
|
51 |
fix the failing CI workflows, and the benchmark will push commits to GitHub, returning the results of the workflow
|
52 |
runs
|
@@ -60,14 +60,14 @@ Each example has the following fields:
|
|
60 |
| Field | Description |
|
61 |
|---------------------|------------------------------------------------------------------------------------------------------------------------------|
|
62 |
| `contributor` | Username of the contributor that committed changes |
|
63 |
-
| `difficulty` | Difficulty of the problem (assessor-based. 0 means that fix requires only the code formatting) |
|
64 |
| `diff` | Contents of the diff between the failed and the successful commits |
|
65 |
| `head_branch` | Name of the original branch that the commit was pushed at |
|
66 |
| `id` | Unique ID of the datapoint |
|
67 |
-
| `language` | Main language of the
|
68 |
| `logs` | List of dicts with keys `log` (logs of the failed job, particular step) and `step_name` (name of the failed step of the job) |
|
69 |
-
| `repo_name` | Name of the original
|
70 |
-
| `repo owner` | Owner of the original
|
71 |
| `sha_fail` | SHA of the failed commit |
|
72 |
| `sha_success` | SHA of the successful commit |
|
73 |
| `workflow` | Contents of the workflow file |
|
|
|
4 |
data_files:
|
5 |
- split: test
|
6 |
path: data/python/*.json
|
|
|
7 |
---
|
8 |
|
9 |
+
# ๐๏ธ Long Code Arena (CI builds repair)
|
10 |
|
11 |
+
This is the benchmark for CI builds repair task as part of the
|
12 |
+
๐๏ธ [Long Code Arena benchmark](https://huggingface.co/spaces/JetBrains-Research/long-code-arena).
|
13 |
+
|
14 |
+
> ๐ ๏ธ CI build repair: given the logs of a failed GitHub Actions workflow and the corresponding repository shapshot, fix the
|
15 |
> repository contents in order to make the workflow pass.
|
16 |
|
17 |
+
All the data is collected from repositories published under permissive licenses (MIT, Apache-2.0, BSD-3-Clause, and BSD-2-Clause). The datapoints can be removed upon request.
|
|
|
18 |
|
19 |
+
To score your model on this dataset, you can use [**CI build repair benchmark**](https://github.com/JetBrains-Research/lca-baselines/tree/main/ci-fixing/ci-fixing-benchmark).
|
|
|
20 |
|
21 |
## How-to
|
22 |
|
|
|
36 |
```
|
37 |
|
38 |
Note that all the data we have is considered to be in the test split.
|
39 |
+
**NOTE**: If you encounter any errors with loading the dataset on Windows, update the `datasets` library (was tested on `datasets==2.16.1`)
|
40 |
|
41 |
|
42 |
## Dataset Structure
|
|
|
44 |
This dataset contains logs of the failed GitHub Action workflows for some commits
|
45 |
followed by the commit that passes the workflow successfully.
|
46 |
|
47 |
+
Note that, unlike other ๐๏ธ Long Code Arena datasets, this dataset does not contain repositories.
|
48 |
|
49 |
+
* Our [CI builds repair benchmark](https://github.com/JetBrains-Research/lca-baselines/tree/main/ci-fixing/ci-fixing-benchmark) clones the necessary repos to the user's local machine. The user should run
|
50 |
their model to
|
51 |
fix the failing CI workflows, and the benchmark will push commits to GitHub, returning the results of the workflow
|
52 |
runs
|
|
|
60 |
| Field | Description |
|
61 |
|---------------------|------------------------------------------------------------------------------------------------------------------------------|
|
62 |
| `contributor` | Username of the contributor that committed changes |
|
63 |
+
| `difficulty` | Difficulty of the problem (assessor-based. 0 means that the fix requires only the code formatting) |
|
64 |
| `diff` | Contents of the diff between the failed and the successful commits |
|
65 |
| `head_branch` | Name of the original branch that the commit was pushed at |
|
66 |
| `id` | Unique ID of the datapoint |
|
67 |
+
| `language` | Main language of the repository |
|
68 |
| `logs` | List of dicts with keys `log` (logs of the failed job, particular step) and `step_name` (name of the failed step of the job) |
|
69 |
+
| `repo_name` | Name of the original repository (second part of the `owner/name` on GitHub) |
|
70 |
+
| `repo owner` | Owner of the original repository (first part of the `owner/name` on GitHub) |
|
71 |
| `sha_fail` | SHA of the failed commit |
|
72 |
| `sha_success` | SHA of the successful commit |
|
73 |
| `workflow` | Contents of the workflow file |
|