TransformersPHP

Appearance

Introduction ​

What is TransformersPHP ​

TransformersPHP is a toolkit for PHP developers to add machine learning magic to their projects easily. You've probably heard about the Python library from Hugging Face, famous for doing awesome stuff with text, like summarizing long articles, translating between languages, and even image and audio related tasks. Transformers PHP brings this capability to the PHP world.

Using Pre-trained Models ​

The core idea behind TransformersPHP is to let you use models that are already trained. "Pre-trained models" are just machine learning models that have been fed and learned from massive amounts of text data. They're ready to go out of the box and can perform a wide range of tasks. With TransformersPHP, these models run directly in your PHP application. That means you don't need to use external services or APIs to process your data. Everything happens locally, on your server.

What's ONNX Runtime? ​

ONNX Runtime might seem like a complex term, but it's essentially a high-performance engine designed for both inferring and accelerating machine learning models. The Open Neural Network Exchange (ONNX) format is at the heart of this engine, serving as a universal format for machine learning models. This means no matter which framework was originally used to train a model — be it PyTorch, TensorFlow, JAX, or even classical machine learning libraries like scikit-learn, LightGBM, XGBoost, etc.—it can be converted to ONNX format. This format can run efficiently across different platforms, including your PHP applications.

Inspired by the Best ​

The development of TransformersPHP was inspired by the Xenova/transformers project, a similar initiative for JavaScript using ONNX runtime too. This shared inspiration means that most models prepared for use with Xenova/transformers, are also compatible with TransformersPHP. It creates a seamless bridge between the machine learning world and PHP development, allowing you to leverage powerful models within your applications.

Quick tour ​

To make things clear, let's compare how you'd do something in Python, PHP (using our library), and Javascript (using a similar library called Xenova). Let's say you want to find out if a piece of text has a positive or negative vibe:

Python (original)PHP (ours)Javascript (Xenova)
python
from transformers import pipeline

# Allocate a pipeline for sentiment-analysis
pipe = pipeline('sentiment-analysis')

out = pipe('I love transformers!')
# [{'label': 'POSITIVE', 'score': 0.999806941}]
php
use function Codewithkyrian\Transformers\Pipelines\pipeline;

// Allocate a pipeline for sentiment-analysis
$pipe = pipeline('sentiment-analysis');

$out = $pipe('I love transformers!');
// [{'label': 'POSITIVE', 'score': 0.999808732}]
javascript
import {pipeline} from '@xenova/transformers';

// Allocate a pipeline for sentiment-analysis
let pipe = await pipeline('sentiment-analysis');

let out = await pipe('I love transformers!');
// [{'label': 'POSITIVE', 'score': 0.999817686}]

You can see how similar it is across languages, making it easier if you're switching between them or learning a new one.

What TransformersPHP is Not ​

While the original HuggingFace Transformers library in Python is a versatile tool supporting both the training of machine learning models and inference (using models to make predictions), TransformersPHP allows only inference. This means that you cannot train new models from scratch, or fine-tune pretrained models using TransformersPHP.