Reimplemtation of Karpathy's micrograd in
- Numpy : Implements a Tensor backed by numpy array with grads for matmul and mean ops
- Torch : Uses torch's tensor and grad. Adds layer and mlp on top
- Triton : Use triton's matmul and add implementation. Inference only.
This was done as an exercise to learn torch/triton.