Automatically collapsing intermediate variables in a python function

I have many hundreds of python functions I'm evaluating for memory usage, and many of them differ simply by the usage of intermediate variable assignments. I would like a way, if possible, to automatically refactor these functions to remove such intermediate variable assignments. This way, the memory differences between the functions would be due to algorithm logic instead of just additional variable assignments.

For example, consider the two functions

def se1(vals1, vals2):
    diff = [(v1 - v2) for v1, v2 in zip(vals1, vals2)]
    diff_sq = [d**2 for d in diff]
    return(sum(diff_sq))

def se2(vals1, vals2):
    return(sum([(x-y)**2 for x,y in zip(vals1, vals2)]))

When I measure memory usage for them using tracemalloc:

import numpy as np
import tracemalloc

listA = np.random.randint(0, 1000, 1000)
listB =  np.random.randint(0, 1000, 1000)

tracemalloc.start()
se1(listA, listB)
current, peak = tracemalloc.get_traced_memory()
tracemalloc.stop()
tracemalloc.reset_peak()
print("Peak memory usage of se1: ", peak)

tracemalloc.start()
se2(listA, listB)
current, peak = tracemalloc.get_traced_memory()
tracemalloc.stop()
tracemalloc.reset_peak()
print("Peak memory usage of se2: ", peak)

the usages are vastly different:

Peak memory usage of se1: 66722

Peak memory usage of se2: 34229

Even though they are producing the same answer using the same logic. I feel like there should be a way to remove intermediate variables using successive substitution within the function body, but I can't think of a general and streamlined way to do it.