Low Accuracy on MNIST Dataset model with custom framework

I'm building a deep learning framework from scratch. Today i just completed linear layer and found out that the model is not learning anything. The weights and bias all are zeros and backpropagation is not working. Here is the github code: [https://github.com/introlix/HeatFlow](https://github.com/introlix/HeatFlow)

I implemented the same model in pytorch and got an accuracy 70% with the same parameters. I only got 10% accuracy on my custom framework.

Never mind the cpp code because first i thought to implement all of …

accuracy code cpp deeplearning framework later mind parameters pytorch them thought