license: mit | |
datasets: | |
- christopher/rosetta-code | |
pipeline_tag: text-classification | |
tags: | |
- code | |
This is a CoreML model for identification of following programming languages: | |
```go, lua, perl, python, apl, shell, c, c#, c++, cobol, lisp, erlang, fortran, groovy, haskell, java, javascript, kotlin, objective-c, pascal, php, powershell, r, ruby, rust, scala, scheme, swift, dart, sql, text, mysql, typescript, ecma, cmake, html, latex, jinja, json, toml, css``` | |
It was trained on a cleaned up and filtered rosetta-code dataset (more precisely: https://huggingface.co/datasets/christopher/rosetta-code, but cleaned up). | |
## ProgrammingLanguageIdentificationV1 | |
First version of PIL model. It was trained on 20 362 data points (including validation, which was picked automatically). | |
Because each programming language has a different number of snippets (lowest: css, ecma, toml (1), highest: go (1110)) its accuracy varies a lot between languages. It's general accuracy is 98,8% for training and validation. | |
Future versions will focus on increasing dataset size. |