Can I train a model with a transformer on a gpu, but load the model and use a cpu?

[thomazmoon]

is possible to train a model with the pipeline ["transformer", "ner"] with a gpu (because of the transformer), but call the model later on using only the cpu later on?

The docs says:

"Transformers are large and powerful neural networks that give you better accuracy, but are harder to deploy in production, as they require a GPU to run effectively."

Because it's referring to deployment, does that mean that using just the prediction will require a gpu?

[thomazmoon]

Answers from previous posts.

1. Originally posted by @polm in Transformer inference on CPU #10543 (comment)

   Transformers should run on CPU, and if they don't that could be a bug, though in most cases they will be too slow to be useful.

2. Originally posted by @spatiebalk in Disabling cuda during inference for transformer-ner model #9925

   I have trained a spacy model with the following components [sentencizer, transformers, ner] in Azure ML Studio using a GPU. When I load and run the model locally I can use the model for inference without any GPU (I don't have one) without any changes.