To run this example, you will have to unzip the MNIST data files in data/.

Saving a trained model by tracing

This tutorial illustrates how to serialize a model trained in Rust as a TorchScript program.

Training and serializing a TorchScript program in Rust.

By running the following command, you can train a model on the MNIST dataset and serialize it as a TorchScript program in Rust.

bash

cargo run --example jit-trace

It will save the model as model.pt.

Loading the model in Python

You can load and run the saved model in Python. test.py is an example Python script to load the model from model.pt and run it on the MNIST dataset.

You can run the script by the following command. Please ensure that torch and torchvision are installed in your environment before running.

bash

python ./examples/jit-trace/test.py