Changes in version 0.1.0 - First release. The CUDA routes of tulpa 0.8.0 (dense and sparse Cholesky factorizations, batched Laplace factorizations, basis products, observation scatter and fused likelihood kernels), the runtime-loaded CUDA layer, the calibrated crossover table and gpu_info() move here from tulpa. The package registers a backend through tulpa's device_backend.h interface when it loads.