Tired of wrestling with platform-specific machine learning model formats? Join Allen Firstenberg and Mark Tucker on Two Voice Devs as they explore ONNX (Open Neural Network Exchange), a game-changing open format built to streamline your ML model deployment workflow. Discover how ONNX empowers you to train models in your preferred framework (PyTorch, TensorFlow, scikit-learn, etc.) and seamlessly execute them across diverse platforms (Windows, Mac, Linux, iOS, Android, Web) using the efficient ONNX Runtime.
In this episode, we delve into:
[00:00:00] Introduction: A warm welcome and a quick overview of the show's agenda.
[00:01:18] What is ONNX?: Unraveling the mysteries of ONNX and its purpose in the ML ecosystem.
[00:02:38] Model Preparation: Understanding how to prepare models for ONNX conversion and the concept of inference.
[00:04:05] Hugging Face Example: A practical demonstration of a BERT model in ONNX format on Hugging Face.
[00:06:00] The Developer's Perspective: Why ONNX matters for developers building applications that leverage ML models.
[00:07:24] ONNX Optimization: How ONNX optimizes models for inference and the trade-offs involved.
[00:08:56] The Cross-Platform Advantage: Breaking free from framework limitations and enabling deployment flexibility.
[00:11:19] ONNX Runtime Introduction: Exploring the ONNX Runtime and its support for various languages and platforms.
[00:14:04] ONNX Runtime Deep Dive: A closer look at the ONNX Runtime website and its features.
[00:15:45] ONNX for Mobile and Web: Extending ONNX's reach to mobile devices and web browsers.
[00:16:56] Conversion Process: Learn how to convert models from different formats to ONNX.
[00:18:08] Performance Considerations: Addressing concerns about performance and speed in ONNX.
[00:19:58] Code Examples: Practical code snippets demonstrating ONNX Runtime usage in JavaScript, Python, and C#.
[00:23:23] ONNX and MLOps: Integrating ONNX into your MLOps pipeline for seamless deployment.
[00:23:42] Netron Tool Introduction: Visualizing ONNX models using the Netron tool.
Whether you're a seasoned data scientist or a developer just beginning your ML journey, this episode provides valuable insights into leveraging ONNX for efficient and cross-platform model deployment. Share your experiences and questions in the comments below!
Thumbnail by Imagen 3 with prompt:
Cartoon ink and paint, with a touch of tech.
Scene: Two podcast hosts, sitting in front of microphones,
smiling and engaging in conversation.
Both hosts are male, caucasian, software developers in their early 50s,
wearing glasses, and are clean shaven.
The host on the left is wearing an olive t-shirt and a brown flat cap.
The host on the right is wearing a light blue polo shirt.
A polished, dark, onyx gemstone, reflecting light and giving it depth.
The gemstone facets should subtly reflect stylized icons of different
operating systems (Windows logo, Apple logo, Android logo, a cloud icon),
hinting at cross-platform compatibility.
Dark, sleek, and mysterious, with the onyx stone as the centerpiece.
The reflected platform icons should be subtle and not overly distracting.
The overall impression should be one of sophisticated power and hidden
potential, alluding to the capabilities of ONNX.
#ONNX #MachineLearning #ML #AI #ArtificialIntelligence #DeepLearning #ModelDeployment #CrossPlatform #PyTorch #TensorFlow #ScikitLearn #MLOps #SoftwareDevelopment #WebDevelopment #MobileDevelopment #JavaScript #Python #CSharp #HuggingFace #Netron
