search

imagej / imagej-tensorflow

ImageJ/TensorFlow integration library
https://imagej.net/libs/tensorflow
Other
25 stars 13 forks source link

Cache loaded models during runtime #13

Closed frauzufall closed 5 years ago

frauzufall commented 6 years ago

This adds a loaded model to a Hashmap to be able to load it faster the next time. We already do the same for graphs and labels, I think this line just got lost?

frauzufall commented 5 years ago

Ok, this oneliner is fixed and ready to be merged. Should I wait for anyone to review it?

ctrueden commented 5 years ago

:+1: LGTM.

As an aside: we might want to consider adding some kind of cache clearing API to the TensorFlowService. As things stand, the only way to free all that memory is to dispose the entire service. IIUC, the loaded models take a substantial amount of memory.

frauzufall commented 5 years ago

Right! I opened an issue https://github.com/imagej/imagej-tensorflow/issues/21.