Building a pipeline from common neural network frameworks to Halide, thus allowing end-to-end functionality on the CGRA. This pipeline utilizes Onnx, an open source format for AI models (ex. PyTorch->Onnx->Halide).