WeGO (Whole Graph Optimizer) is a Vitis AI experimental feature that offers a smooth solution to deploy TensorFlow models on cloud DPU by integrating Vitis AI development kit with TensorFlow framework r1.15. For TensorFlow 2.x, WeGO isn’t available yet.
WeGO automatically performs subgraph partitioning for the Vitis AI quantizer models and applies optimizations and acceleration for the cloud DPU compatible subgraphs. The remaining DPU unsupported parts of the graph are dispatched to TensorFlow for CPU execution. WeGO takes care of the whole graph optimization, compilation, and run-time subgraphs’ dispatch and execution. This process is entirely transparent to the end-users, making it easy to use.
Using WeGO is a straightforward transition from training to inference for model designers. WeGO provides a Python programming interface to deploy the quantized models over the TensorFlow framework. This makes it possible to reuse the Python code (including pre-processing and post-processing) developed during models training with TensorFlow to a maximum. It substantially accelerates the deployment and evaluation of the model over cloud DPUs.
For more information about using WeGO to deploy TensorFlow v1.15 models, see Vitis AI GitHub repo.