Datasets:
Tasks:
Text Generation
Modalities:
Text
Formats:
parquet
Languages:
code
Size:
10K - 100K
ArXiv:
Tags:
code
License:
Update README.md
Browse files
README.md
CHANGED
|
@@ -38,4 +38,17 @@ configs:
|
|
| 38 |
data_files:
|
| 39 |
- split: train
|
| 40 |
path: data/train-*
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
data_files:
|
| 39 |
- split: train
|
| 40 |
path: data/train-*
|
| 41 |
+
task_categories:
|
| 42 |
+
- text-generation
|
| 43 |
+
tags:
|
| 44 |
+
- code
|
| 45 |
---
|
| 46 |
+
|
| 47 |
+
# EditPackFT
|
| 48 |
+
EditPackFT is a dataset built for training LLMs on the task of instructional code editing. The mail columns are:
|
| 49 |
+
1. `before` the code before the edit
|
| 50 |
+
2. `instruction` the instruction to transform the `before` code into the `after` code
|
| 51 |
+
3. `after` the code after the edit
|
| 52 |
+
4. `content` a pre-formatted training window that can be used to train an LLM with prompts in the format of: `<before><instruction><after>`
|
| 53 |
+
|
| 54 |
+
This dataset has been filtered from CommitPackFT. For more detail, [see our paper](https://federico.codes/assets/papers/canitedit.pdf), and our [GitHub repository](https://github.com/nuprl/CanItEdit/tree/main/editpackft).
|