Model Card for Model ID

This is a 🤗 transformers model card that was pre-trained on a dataset built from the JetBrains Kotlin repository on GitHub and uploaded to Hub. This model card was generated automatically.

Model Details

Model Description

The model is a language model pre-trained on Kotlin code from the JetBrains Kotlin repository on GitHub. It is designed to complete and understand code in these programming languages.

Developed by: https://huggingface.co/belka13
Model type: Language model
Language(s) (NLP): Kotlin, Python
Finetuned from model: microsoft/phi-1_5

Model Sources [optional]

Repository: https://github.com/13Belka13/Improving-LLMs-on-underrepresented-Kotlin/tree/main

Uses

Direct Use

The model can be used to augment Kotlin and Python code, finding bugs and style inconsistencies in existing code. It can also help in understanding and documenting code bases.

Out-of-Scope Use

The model is not intended to generate arbitrary text or code in programming languages other than Kotlin and Python. It should also not be used to automatically make changes to production code without human review.

Bias, Risks, and Limitations

The model may have biases and limitations based on code in the JetBrains Kotlin repository. The generated code may not conform to best practices or may contain bugs and vulnerabilities.

Recommendations

Users should be aware of the risks, biases and limitations of the model. Model-generated code should always be tested by experienced developers before use in production. The model is best suited for auxiliary purposes, such as learning Kotlin/Python syntax and style, rather than for writing production code yourself.

Evaluation

Testing Data, Factors & Metrics

Metrics

The following metrics were used to evaluate the model in Kotlin and Python:

Model	Dataset	Accuracy	Perplexity
Default	Kotlin	0.0031	94.78
	Python	0.0067	6.67
Fine-tuned	Kotlin	0.0096	62.82
	Python	0.0180	5.71