krshrimali / Digit-Recognition-MNIST-SVHN-PyTorch-CPP

Implementing CNN for Digit Recognition (MNIST and SVHN dataset) using PyTorch C++ API

Digit-Recognition-MNIST-SVHN

Blog

The blog post on using VGG-Net on MNIST Dataset using PyTorch C++ API: https://krshrimali.github.io/posts/2019/06/introduction-to-pytorch-c-api-mnist-digit-recognition-using-vgg-16-network/.

Implementing CNN for Digit Recognition (MNIST and SVHN dataset) using PyTorch C++ API

Branch - Version 1

1. Using MNIST dataset currently.
2. Installing PyTorch C++ API.
3. Training using PyTorch C++ API.

Note: There may be C10:errors either because of:

1. Incorrect data path. Check your data directory of MNIST dataset.
2. Make sure while using cmake: cmake -DCMAKE_PREFIX_PATH=/absolute_path_to_libtorch/ - it should be absolute path to libtorch.

Process

Note: Prefer using stable version of libtorch. This code doesn't use GPU, as I don't have the resources to test it right now.

1. mkdir build
2. cd build
3. cmake: cmake -DCMAKE_PREFIX_PATH=/home/user/path/to/libtorch/ ..
4. make
5. ./example