Searching for Condition

[LeanderK]

is the equivalent of tf.cond somewhere? I am unable to find it...

[blackgnezdo]

The closest I can think of is select. Can you make it work for your use case? If not, could you share more about your requirements and maybe somebody will have a suggestion for you?

[LeanderK]

maybe it could work, but it's for control flow on the tensor-layer (what I mean is that it doesn't depend on the values of the elements of the tensors). I want to switch between different ways to execute my graph since my training and test environment are different.

[fkm3]

Not sure what you mean. Our select is the same as the python tf.where. Both tf.cond and tf.where act on the values of the predicate at graph execution time, but I think tf.cond has a guarantee that it will only evaluate one branch so that you can use it with mutating ops like tf.assign.