Building a character level language model

• Create a venv and install all required packages from requirements.txt run

python -m venv venv

.\venv\Scripts\activate 
or
venv/bin/activate

pip install -r ./requirements.txt

• To run bigram model to generate names go to bigram folder and run

python run.py 5

we can change the num of names by changing the '5' in cli with number of choice

• To train model we can run

python train.py

This will train the model with data from data\IndianNames.txt and after model is trained it will save the model as onnx in models folder.

I chose onnx because i wanted to create frontend for this as well. And in frontend I used onnx and tfjs to import and run the model and give output. I can make backend api seperately and make a frontend api request but for that i have to write backend and deploy it so to escape from that effort and explore tfjs and onnx i made it this way.Frontend is deployed in vercel. For anyone who is interested this is the site link Indian Name Generator. And this is the repo link Repo

I have hardcoded everything in train.py like the data location. place to store model.learning rate etc etc.I have written the code very badly I will package everything properly in the future but for now this is it.

I am doing this along side a video of andrej karpathy's makemore. This project will be mostly experimenting with things, trying to extend a bit more from the videos and add something based on what i have learned and what i know.