Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,106 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
# For reference on dataset card metadata, see the spec: https://github.com/huggingface/hub-docs/blob/main/datasetcard.md?plain=1
|
3 |
+
# Doc / guide: https://huggingface.co/docs/hub/datasets-cards
|
4 |
+
{}
|
5 |
+
---
|
6 |
+
|
7 |
+
# Dataset Card for Dataset Name
|
8 |
+
|
9 |
+
Python target autocomplete suggestions in the format of conversations for OpenAI's fine-tuning.
|
10 |
+
|
11 |
+
## Dataset Details
|
12 |
+
|
13 |
+
### Dataset Description
|
14 |
+
|
15 |
+
<!-- Provide a longer summary of what this dataset is. -->
|
16 |
+
|
17 |
+
|
18 |
+
|
19 |
+
- **Curated by:** [More Information Needed]
|
20 |
+
- **Funded by [optional]:** [More Information Needed]
|
21 |
+
- **Shared by [optional]:** [More Information Needed]
|
22 |
+
- **Language(s) (NLP):** [More Information Needed]
|
23 |
+
- **License:** [More Information Needed]
|
24 |
+
|
25 |
+
### Dataset Sources [optional]
|
26 |
+
|
27 |
+
The data has been scraped from the following public GitHub repositories on 2023-11-15:
|
28 |
+
|
29 |
+
```
|
30 |
+
https://github.com/numpy/numpy
|
31 |
+
https://github.com/pandas-dev/pandas
|
32 |
+
https://github.com/matplotlib/matplotlib
|
33 |
+
https://github.com/scikit-learn/scikit-learn
|
34 |
+
https://github.com/python-pillow/Pillow
|
35 |
+
https://github.com/psaegert/pmtrendviz
|
36 |
+
https://github.com/psaegert/nli-nec
|
37 |
+
https://github.com/graphdeco-inria/gaussian-splatting
|
38 |
+
https://github.com/lllyasviel/ControlNet
|
39 |
+
https://github.com/maltfield/awesome-lemmy-instances
|
40 |
+
https://github.com/Aleph-Alpha/aleph-alpha-client
|
41 |
+
https://github.com/MaartenGr/BERTopic
|
42 |
+
https://github.com/MilesCranmer/PySR
|
43 |
+
https://github.com/AUTOMATIC1111/stable-diffusion-webui
|
44 |
+
https://github.com/microsoft/Codex-CLI
|
45 |
+
https://github.com/dropbox/hydra
|
46 |
+
https://github.com/HLearning/unet_keras
|
47 |
+
https://github.com/hmason/ml_class
|
48 |
+
https://github.com/django/django
|
49 |
+
https://github.com/encode/django-rest-framework
|
50 |
+
https://github.com/pallets/flask
|
51 |
+
https://github.com/postmanlabs/httpbin
|
52 |
+
https://github.com/jakevdp/PythonDataScienceHandbook
|
53 |
+
https://github.com/donnemartin/data-science-ipython-notebooks
|
54 |
+
https://github.com/tensorflow/tensorflow
|
55 |
+
```
|
56 |
+
|
57 |
+
## Uses
|
58 |
+
|
59 |
+
This dataset is intended to be used for fine-tuning of GPT-3.5-Turbo via OpenAI's fine-tuning API.
|
60 |
+
|
61 |
+
## Dataset Structure
|
62 |
+
|
63 |
+
`train_completions.jsonl` includes a list of 100 target conversations. Each conversation is structured as follows:
|
64 |
+
|
65 |
+
```json
|
66 |
+
[
|
67 |
+
{"role": "system", "content": <system prompt>},
|
68 |
+
{"role": "user", "content": <first half of the code>},
|
69 |
+
{"role": "assistant", "content": <small target completion from ground truth>}
|
70 |
+
]
|
71 |
+
```
|
72 |
+
|
73 |
+
## Dataset Creation
|
74 |
+
|
75 |
+
### Curation Rationale
|
76 |
+
|
77 |
+
This dataset was created to fine-tune GPT-3.5-Turbo to provide more reliably formatted autocomplete suggestions for python code.
|
78 |
+
|
79 |
+
#### Data Collection and Processing
|
80 |
+
|
81 |
+
We scraped a mix of 25 GitHub repositories related to python and sampled 4 random python files weighted by their length.
|
82 |
+
The files were split in 2 parts at a uniformly random point in each file.
|
83 |
+
Next, if the input was longer than 10k tokens, we truncated the input file to 250 to 10000 tokens from the beginning.
|
84 |
+
The output was manually truncated to reasonably short code completions.
|
85 |
+
|
86 |
+
#### Who are the source data producers?
|
87 |
+
|
88 |
+
Paul Saegert
|
89 |
+
|
90 |
+
[More Information Needed]
|
91 |
+
|
92 |
+
#### Personal and Sensitive Information
|
93 |
+
|
94 |
+
This dataset may contain personal or sentitive information.
|
95 |
+
|
96 |
+
## Bias, Risks, and Limitations
|
97 |
+
|
98 |
+
The dataset contains only python code from the most popular, trending, or personal projects. It may be biased towards a particular style of code.
|
99 |
+
|
100 |
+
## Citation [optional]
|
101 |
+
|
102 |
+
<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
|
103 |
+
|
104 |
+
**BibTeX:**
|
105 |
+
|
106 |
+
[More Information Needed]
|