Evaluate my newly trained model but output the activations at an intermediate layer

You could do something like:

    BrainScriptNetworkBuilder = {
        network = BS.Network.Load ("pathToYourModel")
        outputNodes = (network.yourIntermediateNode)
    }

If yourIntermediateNode contains periods (.), please replace them with underscores (_). If your network contained dynamic axes, you might have to redefine them.