A fast, pure-Python reimplementation of diffvg in torch. Achieves numerical parity with the original C++/CUDA implementation while being significantly simpler and faster for batched workloads.
Latest commits.
Builders behind this project.