Right now it's possible to use tf.keras.callbacks.ModelCheckpoint to save (keras) model weights during training under NengoDL, and then use sim.keras_model.load_weights(...) afterwards. However it might be nice to have our own version of this that uses the native save_params and load_params directly.
I briefly looked at the ModelCheckpoint code and after a quick scan it looks like it should be pretty reasonable to subclass.
Right now it's possible to use
tf.keras.callbacks.ModelCheckpointto save (keras) model weights during training under NengoDL, and then usesim.keras_model.load_weights(...)afterwards. However it might be nice to have our own version of this that uses the nativesave_paramsandload_paramsdirectly.I briefly looked at the
ModelCheckpointcode and after a quick scan it looks like it should be pretty reasonable to subclass.