FuncHub 0.1.4

A library containing common functions developers used to write repetitevly

Utility Library

A collection of utility functions for file operations, text processing, AI, and machine learning, designed to streamline common tasks in PyTorch projects.

Features

• Load and save JSON and YAML files.
• Text processing utilities including tokenization and normalization.
• Device management for PyTorch (CPU, CUDA, or MPS).
• Similarity computation for embeddings using cosine similarity.
• Utility functions for data manipulation such as moving averages, safe division, and flattening dictionaries.

Installation

You can install the library via pip:

pip install FuncHub

Usage

File Operations

Load YAML File

from pytorch_utility_library import open_yaml

config = open_yaml('config.yaml', key='model')

Load JSON File

from pytorch_utility_library import open_json

data = open_json('data.json')

Save Data to JSON

from pytorch_utility_library import dump_to_json

dump_to_json('output.json', data)

Save Text to File

from pytorch_utility_library import dump_to_text

dump_to_text('Hello, World!', 'output/hello.txt')

Text Processing

Tokenize Text

from pytorch_utility_library import tokenize

tokens = tokenize("Hello, how are you?")

Normalize Text

from pytorch_utility_library import normalize_text

normalized = normalize_text("  Hello,   How Are You?  ")

AI and Machine Learning

Get Device

from pytorch_utility_library import get_device

device = get_device()
print(f"Using device: {device}")

Compute Cosine Similarity

from pytorch_utility_library import compute_similarity

similarity = compute_similarity(embedding1, embedding2)

Utility Functions

Calculate Moving Average

from pytorch_utility_library import moving_average

averages = moving_average([1, 2, 3, 4, 5], window_size=3)

Flatten a Nested Dictionary

from pytorch_utility_library import flatten_dict

flat_dict = flatten_dict({'a': {'b': 1, 'c': 2}, 'd': 3})

Logging

The library uses a custom logger for error and information logging. Make sure to configure the logger in your application as needed.

Contributing

Contributions are welcome!

License

This project is licensed under the MIT License - see the LICENSE file for details.