DL Framework
Building a compact PyTorch-style deep learning framework to understand tensors, autograd, modules, dispatch, and native performance paths.
Building a compact PyTorch-style deep learning framework to understand tensors, autograd, modules, dispatch, and native performance paths.
Python Tensor Library Tensor object, broadcasting, indexing, reductions, and NumPy-backed operations
Autograd Engine Dynamic computation graph, reverse-mode differentiation, and gradient checks
Neural Network Modules Parameter management, layers, losses, optimizers, and training loops
C++ Backend Native tensor kernels, Python bindings, memory layout, and extension interfaces