search

hughperkins / DeepCL

OpenCL library to train deep convolutional neural networks
Mozilla Public License 2.0
867 stars 199 forks source link

Hook for model serialization #26

Closed RyanHope closed 8 years ago

RyanHope commented 9 years ago

I've modified your qlearning example to play tetris and all is fine and dandy but it would be nice to save the net somehow so that I can resume training or use the trained net in another application. Are there any entry points in the current API for doing this or do I need to write something my self?

hughperkins commented 9 years ago

Hi Ryan, you can use WeightsPersister class to persist the weights. You can see an example of loading the weights at https://github.com/hughperkins/DeepCL/blob/master/src/main/train.cpp#L329 , and saving the weights at https://github.com/hughperkins/DeepCL/blob/master/src/main/train.cpp#L386

hughperkins commented 8 years ago

Closing this since old. If I haven't quite addressed your question, please go ahead and reopen :-)