Updating custom translation model and Google base-model doubt

[elgonher]

Hello,

When we train a custom translation model using Cloud Translation (or before AutoML), what's the best approach regarding the re-training of those models so they keep up with Google's newest updates? For example, if a custom model was trained 2 years ago, it won't have the latest updates, right? I guess that, in the time of training, it uses the most updated google base model, but then as time passes it will become outdated, is that correct?

Thanks in advance.

[McMaco]

Hello elgonher,

Welcome to Google Cloud Community!

The best approach I can recommend regarding the re-training of those models so they keep up with Google's newest updates is to evaluate models. To improve the quality of your model, consider adding more (and more diverse) training segment pairs. After you adjust your dataset, train a new model by using the improved dataset.

In addition, you can choose to upgrade some or all of your resources. When you upgrade a dataset, any models that are associated with that dataset are also automatically upgraded. Only models without an underlying dataset (like in cases where the associated dataset was deleted) can be manually upgraded on their own.

You can use these step by step guides for upgrading resources in the Google Cloud console to upgrade existing AutoML resources to Cloud Translation resources.

I hope the above information is helpful.