Update README.md
Browse files
README.md
CHANGED
|
@@ -47,6 +47,25 @@ configs:
|
|
| 47 |
# Dataset Summary
|
| 48 |
The dataset contains 25,000 Kotlin code samples selected from the [KStack](https://huggingface.co/datasets/JetBrains/KStack) dataset. The selection is performed based on the value of the code for learning algorithmic concepts in Kotlin. In total, the dataset contains about 23M [CodeLlama-7b](https://huggingface.co/codellama/CodeLlama-7b-hf) tokens (vocab size 32,016).
|
| 49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
# Dataset Collection
|
| 51 |
The filtering from [KStack](https://huggingface.co/datasets/JetBrains/KStack) is performed using zero-shot quality estimation based on [Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2). The model is prompted to determine which of two files has higher "educational value for learning algorithms in Kotlin". The results of the comparisons are averaged and used to train a binary classifier based on [CodeT5p-220m](https://huggingface.co/Salesforce/codet5p-220m). The binary classifier is then applied to the entire KStack to obtain scores for each sample in the dataset. The log-probability of the classifier prediction is used as a criterion of the selection.
|
| 52 |
|
|
|
|
| 47 |
# Dataset Summary
|
| 48 |
The dataset contains 25,000 Kotlin code samples selected from the [KStack](https://huggingface.co/datasets/JetBrains/KStack) dataset. The selection is performed based on the value of the code for learning algorithmic concepts in Kotlin. In total, the dataset contains about 23M [CodeLlama-7b](https://huggingface.co/codellama/CodeLlama-7b-hf) tokens (vocab size 32,016).
|
| 49 |
|
| 50 |
+
## Column description
|
| 51 |
+
The dataset contains the following columns:
|
| 52 |
+
|
| 53 |
+
- `size` — size of the file in bytes
|
| 54 |
+
- `content` — text (content) of the file after removing personal identifiable information
|
| 55 |
+
- `repo_id` — GitHub ID of the repository
|
| 56 |
+
- `path` — path to a file
|
| 57 |
+
- `owner` — repo owner on GitHub
|
| 58 |
+
- `name` — repo name on GitHub
|
| 59 |
+
- `commit_sha` — hash of the commit, from which the revision of the file is taken
|
| 60 |
+
- `stars` — number of stars in the repo at the moment of collection
|
| 61 |
+
- `forks` — number of forks in the repo at the moment of collection
|
| 62 |
+
- `issues` — number of issues in the repo at the moment of collection
|
| 63 |
+
- `is_fork` — `true` if the repo is a fork or not as defined by GitHub
|
| 64 |
+
- `main_language` — main language of the repo as defined by GitHub
|
| 65 |
+
- `languages_distribution` — JSON with the distribution of languages by size in bytes in the repo
|
| 66 |
+
- `license` - the name of the permissive license
|
| 67 |
+
|
| 68 |
+
|
| 69 |
# Dataset Collection
|
| 70 |
The filtering from [KStack](https://huggingface.co/datasets/JetBrains/KStack) is performed using zero-shot quality estimation based on [Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2). The model is prompted to determine which of two files has higher "educational value for learning algorithms in Kotlin". The results of the comparisons are averaged and used to train a binary classifier based on [CodeT5p-220m](https://huggingface.co/Salesforce/codet5p-220m). The binary classifier is then applied to the entire KStack to obtain scores for each sample in the dataset. The log-probability of the classifier prediction is used as a criterion of the selection.
|
| 71 |
|