Performance issue in lime/scripts/lipo_gru/ht_lipo_gru.py

[DLPerf]

Hello! Our static bug checker has found a performance issue in lime/scripts/lipo_gru/ht_lipo_gru.py: init is repeatedly called in a for loop, but there are several tf.function decorated functions call, call and call defined and called in init.

In that case, when init is called in a loop, the function call will create a new graph every time, and that can trigger tf.function retracing warning.

Similar issues in: lime/scripts/lipo_gru/ht_lipo_ind_gru.py, lime/scripts/elf_with_wbo/ht_elf_with_wbo.py, lime/scripts/partial_charge/ht_charge.py and lime/scripts/partial_charge/ht_charge_fast.py.

Here is the tensorflow document to support it.

Briefly, for better efficiency, it's better to use:

@tf.function
def inner():
    pass

def outer():
    inner()  

than:

def outer():
    @tf.function
    def inner():
        pass
    inner()

Looking forward to your reply.

[DLPerf]

There are some variables in class passed to the function, the code may be more complex if changes are made. Do you have any idea? @dotsdl @jchodera @bas-rustenburg @jaimergp