This repository contains code that trains four deep and shallow models and backend to serve them. The code for the training is found in the notebook inside the notebook folder.
I used data from OkCupid to train four models where the target is "status" column of the data. Status being single, married, in relationship, etc.
You can find it here ---> https://willsh3b33.xyz
There are two shallow models in this model. One's a gradient boost using XGBoost, and one's an OVR SVM using Libsvm through Sklearn.
There's a simple feed-forward model that takes the same input features as the shallow models. This model had a high accuracy, and was constructed using Keras.
There's an overall of 15 preprocessors. You can find a list of them in the .env file that I have linked below.
I have fine-tuned BERT on the 9 essays of this dataset. All 9 essays at once. But the latent space is a free-for-all, just write everything you feel like that represents you!
It's the BERT tokenizer that comes with the transformers library.
I haven't included the models in this repositry because it's a bad practice to do so. You can download my .env file from here, and put it in the root folder. Then you can launch app.py, and use Postman to send requests.
Link:
https://cdn.discordapp.com/attachments/797919965264085092/889497440179343360/env
NOTE: Name of the file should be env, not .env!
The frontend is coming soon!