Why Epochs take longer as learning proceeds?
Pergunta
I am training a deep learning model in Tensorflow
on GPU
(Amazon AWS) and what I observe is that in the beginning each epoch takes only less than a second
but let's say after 80 epochs one epoch takes more than 8 minutes
.
What could be a possible reason?
Nenhuma solução correta
Licenciado em: CC-BY-SA com atribuição
Não afiliado a datascience.stackexchange