Worked on the pytorch/pytorch repository to deliver export functionality for the Scaled Dot-Product Attention (SDPA) operation to the ONNX Attention operator, targeting enhanced interoperability in ONNX-based workflows. The implementation, developed in Python using PyTorch and ONNX, included comprehensive unit tests to validate the export path and ensure compatibility with ONNX opset version 23. This work addressed the need for seamless production deployment by enabling PyTorch SDPA models to be exported and run efficiently within ONNX runtimes. The focus on robust test coverage and standards compliance contributed to smoother integration and inference for machine learning and deep learning applications.