Helpers & Utilities

Module: trueml.helpers

Benchmarking, profiling, and data generation utilities to facilitate machine learning experiments and performance testing.

---

timeit

@trueml.helpers.timeit
def func(*args, **kwargs)

A decorator that measures and prints the execution time of a function. The number of execution iterations is dynamically controlled by the I environment variable.

Parameter	Type	Default	Description
func	callable	—	The function to be timed.

Environment Variable I: - I >= 1 (default 10): Runs the function I times sequentially, printing the execution time (in seconds) for each run. - I = -1: Runs the function in an infinite loop, continuously printing the execution time alongside the function's name.

Example:

import os
import numpy as np
from trueml.helpers import timeit
from trueml.linalg import matmul

os.environ["I"] = "3"

@timeit
def benchmark():
    A = np.random.randn(50, 50)
    B = np.random.randn(50, 50)
    matmul(A, B)

benchmark()
# 0.0412
# 0.0401
# 0.0405

---

memprofile

@trueml.helpers.memprofile
def func(*args, **kwargs)

A decorator that profiles the memory allocations of a function using Python's built-in tracemalloc library.

Parameter	Type	Default	Description
func	callable	—	The function to be memory-profiled.

Output: Prints the top 10 memory-allocating lines of code executed during the function call.

Example:

from trueml.helpers import memprofile
import numpy as np

@memprofile
def allocate():
    X = np.zeros((10000, 10000))
    return X

allocate()
# <tracemalloc statistics output>

---

generate

trueml.helpers.generate(lower: int = 1, upper: int = 100, size: tuple = (1, 1)) -> np.ndarray

Generates a random integer matrix drawn from a discrete uniform distribution.

Parameter	Type	Default	Description
lower	int	1	The inclusive lower bound of the generated integers.
upper	int	100	The exclusive upper bound of the generated integers.
size	tuple	(1, 1)	The shape of the output matrix.

Returns: A NumPy array of the specified size populated with random integers.

Example:

from trueml.helpers import generate

X = generate(lower=0, upper=5, size=(2, 3))
print(X)
# [[1, 4, 0],
#  [2, 2, 3]]

See Also

• linalg — Mathematical primitives that can be benchmarked using these tools.