A simple example of exporting a transformer model with Python, then loading it into tract to make predictions.

To Use

First export the pre-trained transformer model using Python and PyTorch

python export.py

the exported model and tokenizer are saved in ./albert. Then load the model into tract and make prediction.

cargo run --release

The output for input sentence "Paris is the [MASK] of France" should look like

Result: Some("▁capital")