Hi I notice the aim of the function(CondensingLinear) is mentioned in this answer https://github.com/ShichenLiu/CondenseNet/issues/6#issuecomment-351598403,
Does CondensingLinear in layers.py only be used at convert_model phase, not in the training phase?
Therefore the evaluation error on converted model will be higher than the model in the
training phase, right? It is like directly pruning out 50% of the last fully connected layers
without finetuning.
Yes, that is correct. However, it does not affect the test error significantly. We reported in the paper the actual eval error rate instead of training phase eval error rate.
Hi I notice the aim of the function(CondensingLinear) is mentioned in this answer https://github.com/ShichenLiu/CondenseNet/issues/6#issuecomment-351598403, Does CondensingLinear in layers.py only be used at convert_model phase, not in the training phase? Therefore the evaluation error on converted model will be higher than the model in the training phase, right? It is like directly pruning out 50% of the last fully connected layers without finetuning.