Загрузка...

ML inference in PHP by example leverage ONNX and Transformers on Symfony SymfonyCon 2024

Curious if you can run powerful AI and Machine Learning models directly with PHP? The answer is yes! In this 30-minute talk, we dive into how you can leverage the power of AI in your PHP applications without relying on Python or external APIs.

Learn about the core concepts of Transformer models, see practical live demos, and discover the libraries that make it all possible. From text and image classification to generating text with small language models, this session covers everything you need to get started.

Timestamps:
00:00 - Introduction
00:22 - Start of the Talk
01:01 - What are AI Transformers? (The good kind, not the movie!)
03:27 - The Concept of ""Self-Attention"" in Models
05:36 - Demo: How Transformers Understand Context
08:04 - Bringing AI to PHP with the Transformers PHP Library
10:00 - Use Case 1: Text Classification (Analyzing Amazon Reviews)
12:00 - Demo: Scoring Product Reviews as Positive or Negative
13:33 - Use Case 2: Image Classification
14:44 - Demo: The Classic ""Hot Dog or Not Hot Dog"" App
16:00 - Use Case 3: LLM Text Generation with Small Models
17:30 - Demo: Answering the age-old question, ""Is a taco a sandwich?""
18:34 - How It Works: The Tech Behind Transformers PHP
18:47 - The PHP FFI Extension
19:40 - ONNX (Open Neural Network Exchange)
21:34 - Where to Find Models (Hugging Face & Small ML)
23:19 - Current Limitations: No GPU Support (and a note on Apple Silicon)
25:38 - Options for Running Larger LLMs
27:14 - Q&A Session
27:40 - Can this library use APIs like ChatGPT?
28:25 - Can you train or fine-tune models with this?
29:44 - Are the models language-specific?
30:43 - Are the library's tasks extendable?
32:26 - How do you find the best model for your use case?
33:17 - How do you test the results and accuracy of a model?

Видео ML inference in PHP by example leverage ONNX and Transformers on Symfony SymfonyCon 2024 канала Upsun
Яндекс.Метрика
Все заметки Новая заметка Страницу в заметки
Страницу в закладки Мои закладки
На информационно-развлекательном портале SALDA.WS применяются cookie-файлы. Нажимая кнопку Принять, вы подтверждаете свое согласие на их использование.
О CookiesНапомнить позжеПринять