Remember that this is the “Functional API” now, so you need to instantiate the layer function (specify the parameters you want) and then invoke the resulting function with an actual input tensor (the “(F)” that Tom was referring to in the Dense case). There should be two separate sets of parentheses on each line involving a layer. Did you actually read the thread I linked on our earlier discussion about the Sequential and Functional APIs? If you are trying to save time by “winging it” here and not actually reading the information, that may not actually end up being a net savings of time. For any TF function, you can also find the documentation by searching “tensorflow flatten” e.g. …