scratch cnn

This is an from scratch implementation of cnn.

cnn_manual.py contains scratch implementation of a cnn layer and cnn_part2.py contains scratch implementation of cnn net.

I don't recommend running both since it would take eternity to run since they are not optimized. But are good for learning purpose.

You may run cnn_using_modules.py it shows how to train a cnn network and how to save it. The model is trained on cifar-10 dataset.

How to run this code

git clone https://github.com/KrishnaAgarwal1308/scratch_cnn.git
cd scratch_cnn
pip install torch, torchvision, numpy
python cnn_using_modules.py

scratch cnn

This is an from scratch implementation of cnn.

cnn_manual.py contains scratch implementation of a cnn layer and cnn_part2.py contains scratch implementation of cnn net.

I don't recommend running both since it would take eternity to run since they are not optimized. But are good for learning purpose.

You may run cnn_using_modules.py it shows how to train a cnn network and how to save it. The model is trained on cifar-10 dataset.

you can now run alex-net and ViT from this repo and look at it's implemenation from scratch. Both of which could be selected through run_models.py and could be run individually.

How to run this code

git clone https://github.com/KrishnaAgarwal1308/scratch_cnn.git
cd scratch_cnn
pip install torch, torchvision, numpy
python cnn_using_modules.py

To run visual transformers

python run_models.py

Note: while running ViT make sure to shrink down the size of the image to something like 64 or 128 or else it won't fit in most commercial gpu's vram.