QuickBench 0.0.121

A lightweight python package to time python functions and blocks, and bench machine learning models

QuickBench

The Essential Performance Toolkit for Python.

A [Minerva AI] Project.

A lightweight, zero-dependency Python utility for timing code execution. QuickBench provides both a decorator and a context manager which can easily measure how long functions or code blocks take to run.

QuickBench is the all-in-one performance toolkit for Python developers. Whether you are optimizing a for loop, comparing Machine Learning models, or evaluating LLM latency, QuickBench handles the metrics so you can focus on the code.

Why QuickBench?

• Universal Timer: Benchmark functions or blocks of code with a simple decorator.
• ML Benchmarking: Automatically compare sklearn, xgboost, etc. (Accuracy, F1, RMSE).
• LLM Benchmarking: Measure Token/sec and Latency for GPT, Llama, or Claude wrappers.
• pandas Output: All results return clean DataFrames ready for analysis.

---

Features

• Decorator Support: Time entire functions with a single line (@monitor).
• Context Manager: Time specific blocks of code inside a function (with monitor():).
• Human Readable: Automatically formats output to ns, µs, ms, or s.
• Zero Dependencies: Pure Python, no heavy libraries required.

Installation

pip install QuickBench

Tutorial

Tutorial & Usage

1. The Universal Timer (@monitor) Stop writing start = time.time() manually. Use the decorator to time functions, or the context manager to time specific blocks.

A. Function Decorator Use this to measure how long a specific function takes to run.

from QuickBench import monitor
import time

@monitor
def heavy_processing():
    # Simulates a slow task
    time.sleep(1.5)
    return "Done"

heavy_processing()
# Output: [heavy_processing] finished in 1.50 s

B. Context Manager (Block Timer) Use this when you only want to time a few lines of code inside a larger function.

from QuickBench import monitor
import time

def data_pipeline():
    print("Loading data...")
    
    with monitor(label="Data Cleaning"):
        # Time only this specific part
        time.sleep(0.3)
    
    print("Pipeline finished.")

2. The Auto-ML Benchmarker QuickBench automatically detects if your problem is Classification or Regression and calculates the correct metrics (Accuracy/F1 vs. MSE/R2).

Step 1: Define your models You can use any model that follows the scikit-learn API (has .predict()).

from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier

models = {
    "Logistic": LogisticRegression(),
    "Decision Tree": DecisionTreeClassifier(),
    "Random Forest": RandomForestClassifier()
}

Step 2: Run the Benchmark Pass your dictionary of models along with your test data.

from QuickBench.ai import MLBencher

# Assuming you have X_test and y_test ready
bencher = MLBencher(models, X_test, y_test)
results = bencher.run()

print(results)

Output (Returns a Pandas DataFrame):

Model Type	Name	Primary Score	F1 Score	Latency (s)
Classification	Random Forest	0.9820	0.9815	0.1204
Classification	Decision Tree	0.9650	0.9648	0.0052
Classification	Logistic	0.9400	0.9390	0.0021

Contributing & Issues

We would absolutely love your contribution, since that is what makes us keep going. Found a bug? Want to add a feature?

Report Issues: GitHub Issues Page

Source Code: GitHub Repository

Branding

QuickBench is proudly developed and maintained by Minerva AI. Empowering developers with wisdom and speed.