This project is a powerful fine tune model crafted in google colab to detect and fix all kinds of bugs of a python code in realtime. This is integrated in a VS code extenstion through felix-detect-fix. It's fast with it's GPU/CPU based workflow and compatible with all the Operating System. thanks to the new codebert-base model and deepseek.
If you want to use this model locally then you should have installed python 3+ on your machine. And for model usage if you have a good GPU it will be cherry on the top. But if you don't have any GPU please make sure to use google colab or kaggle notebook or other services. Alternatively You can use any popular cloud services such that AWS, GCP, Azure etc.
Instructions
The both bug detecting and fixing model is available on the Huggingface-bug-detection and Huggingface-bug-fixing if you want to install them you are good to go.
- Open the
usingModel.ipynband test this model. - Check the accuracy score, F1, Precesion and other metrics for better understanding.
- Next, Check for the graphical section to understand confusion matrix, box-plot, histogram etc.
If you want to explore the extension please refer to this repository https://github.com/felixoder/felix-detect-fix
Instructions How to use this
Open the extension tab in your VS code and find this felix-detect-fix and install this.
- Now create a python file having extension of
.ipynb. - Write some code and press
Ctrl/cmd + shift + pand typeDetect Bug. after running this wait for couple of second and if there is a bug you can see a pop up and if you want to clickFix Bug. - Next, Click the
Fix Bug[you can do this either by running theCtrl/cmd + shift + pand typeDetect Bugor the steps in step 2 and do fix likewise]
Here are some useful documentation links:
- Getting started guide:
- VS code extension library: https://marketplace.visualstudio.com/items?itemName=DebayanGhosh.felix-detect-fix
we have evaluated the evaluation matrics of the fine tune model please check the following graphs
and from the evaluation matrics our scores are follows -
Accuracy: 80.00% Precision: 1.00 Recall: 0.60 F1 Score: 0.75
- see how it is working on Windows
To give feedback, ask a question or make a feature request, you can either use the Github Discussions.
Bugs are logged using the github issue system. To report a bug, simply open a new issue.
All contributions are welcomed.
For new nodes, check out this documentation page on how to create a new shader-based node. Once you have it working, prepare a pull request against this repository.
In case you have any questions about a feature you want to develop of something you're not sure how to do, you can still create a draft pull request to discuss the implementation details.
You can install the extension and detect the bug like
Use the model in google colab like this
Please check out my documentation page
@source - The idea and the project is done from my industrial training of Intel Unnati where I am selected among 50 students in my college.
Debayan Ghosh @felixoder uni student code - BWU/BTA/22/157