This dataset is a parallel code corpus mapping standard PyTorch tensor operations to their highly optimized, custom Triton GPU kernel implementations.
Each sample is a JSON object containing:
pytorch: Python string containing the reference PyTorch function.
triton: Python string containing the equivalent optimized Triton kernel (@triton.jit) and its Python launcher function.
