Question about the clusters

Hi, I am very interested in your work.

I have a question about the offline clustering. It seems that you save a cluster dict for each data. Does this mean that this method must sample all eval data offline and obtain high accuracy. What should I do if the eval data cannot be sampled?

Thank you so much.