TTA not using GPU

I am trying to perform a TTA
model = models.resnet50
learn = create_cnn(data, model, metrics=error_rate)
preds_t,y_t = learn.TTA(ds_type=learn.data.test_ds)

I found all the 8 thread (4 cores) CPU to be 100% while GPU is idle.
It’s normal or anything wrong?

Platform: GCP
fastai version: 1.0.38.dev0

Thanks.