i am trying to use google cloud TPU for inferring model.
in GPU i used
del pipeline
torch.cuda.empty_cache()
to free memory for new model.
but how can i free TPU memory?
i am trying to use google cloud TPU for inferring model.
in GPU i used
del pipeline
torch.cuda.empty_cache()
to free memory for new model.
but how can i free TPU memory?