This is an experimental project in the very early stage.
The goal of this project is to explore the feasibility of providing AI training capability for WasmEdge Runtime. It is built up as a plugin. The rudimentary architecture of this project is shown as below:
In the current stage, three examples are provided: one is PyTorch model, the others are TensorFlow models. The models are defined with PyTorch/TensorFlow Python APIs, while the training is done on WasmEdge Runtime via the wasmedge-nn-training
plugin.
-
Example 2: Training a custom model (TensorFlow)
-
Example 3: Training Resnet model (PyTorch)