- To run examples:
PYTHONPATH="." python3 ./examples/or_gate.py
- For MNIST: check
./mnist.ipynb
- A functional tensor library
- Passing Backward passes with tests
- MLP working
- Basic CNN
- CIFAR-10/MNIST working with the lib (MNIST working)
- Add a different backend than numpy
- Current plan is to convert the autograd graph into a set of basic ops and compile that to C/Metal/CUDA(not decided yet) kernels
- Yes this is inspired from tinygrad
Warning
Solely for teaching myself how pytorch works