Class Central is learner-supported. When you buy through links on our site, we may earn an affiliate commission.

YouTube

GPT OSS Release - Inference and Fine-Tuning

Trelis Research via YouTube

Overview

Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn to work with OpenAI's newly released open-source GPT models through comprehensive hands-on implementation covering inference, fine-tuning, and advanced optimization techniques. Explore the technical specifications and performance characteristics of the GPT OSS 20B parameter model, including its mixture of experts architecture and quantization capabilities using the innovative MXFP8 data type. Master running inference on these models using HuggingFace transformers, then dive deep into fine-tuning workflows with practical code examples and debugging strategies. Discover advanced fine-tuning techniques including TensorBoard integration for training visualization, validation split creation for model evaluation, and chat template inspection for proper data formatting. Gain insights into training parameter optimization, model preparation best practices, and development workflow management including SSH connections and version control integration. Access bonus content exploring the relationship between open-source and proprietary model architectures, providing strategic context for model selection and deployment decisions in production environments.

Syllabus

00:00 Introduction to OpenAI's OSS Models
00:16 Model Specifications and Performance
02:36 Quantization and Mixture of Experts
04:00 MXFP8 Data Type
08:46 Running Inference on OSS Models
13:01 Fine-Tuning GPT OSS Models with transformers
24:43 Exploring Track IO Feature
26:33 ADVANCED fine-tuning tricks
27:37 Using TensorBoard for Logging
29:03 Creating a Validation Split
30:12 Inspecting the Chat Template
31:30 Preparing the Model for Training
31:47 Training Parameters and Debugging
46:41 Connecting to the Pod via SSH to git commit
49:06 Final Thoughts on OpenAI OSS Models
49:52 BONUS: Insights on Proprietary Models

Taught by

Trelis Research

Reviews

Start your review of GPT OSS Release - Inference and Fine-Tuning

Never Stop Learning.

Get personalized course recommendations, track subjects and courses with reminders, and more.

Someone learning on their laptop while sitting on the floor.