The ONNX interchange format facilitates the migration and exchange of artificial intelligence algorithms in different computing frameworks and computing hardware. At present, many artificial intelligence development platforms using domestic chips support loading algorithm model files in ONNX format and running algorithms.

Converting the model developed by TensorLayerX into an ONNX format file can further improve the compatibility with home-made AI products. At the same time, the algorithm developed by TensorLayerX on the national AI platform will be converted to ONNX format and can be used by other mainstream frameworks, which improves the compatibility between the national AI ecology and the external environment.

GammaGL, GAMMA LAB and Pengcheng Laboratory provide more than 20 mainstream classical models with the latest open-source graph neural network algorithm library

In addition, GammaGL is very similar to the mainstream PyTorch Geometric (PyG) interface design. You can develop it as a PyG that supports TensorFlow, Paddle, MindSpore, and switch back-end deep learning frameworks with one click.

At the same time, we provide more than twenty mainstream classical models for you to reuse.