Set up the Trainer

Your model will replace complex phrases with simpler, more accessible paraphrases to improve the readability of translations. For example, it can simplify the original phrase "The inclement weather conditions precipitated the postponement of the outdoor event" with "The bad weather caused the outdoor event to be delayed." Build the Trainer to prepare training your language translation service! The exercise will take some time to run with the call to trainer.train().

Some data has been pre-loaded:

model is a Transformer model
dataset contains the MRPC dataset of sentence paraphrases
compute_metrics function returns accuracy and F1 score
You've defined training_args in a prior exercise

Deze oefening maakt deel uit van de cursus

Efficient AI Model Training with PyTorch

Cursus bekijken

Oefeninstructies

Pass in the model to the Trainer() class.
Input the training arguments in the Trainer() class.
Pass in a function to compute metrics to the Trainer() class.
Print the device that the trainer chooses.

Praktische interactieve oefening

Probeer deze oefening eens door deze voorbeeldcode in te vullen.

trainer = Trainer(
    # Pass in the model
    ____=____,
    # Input the training arguments
    ____=____,
    train_dataset=dataset["train"],
    eval_dataset=dataset["validation"],
    # Pass in a function to compute metrics
    compute_metrics=____,
)

trainer.train()

# Print the device that the trainer chooses
print(____.args.____)

Code bewerken en uitvoeren