Description

Fine-Tuning and Optimizing Small Language Models. This course is a comprehensive and practical program that teaches how to fine-tune, optimize, align, and compress small language models for efficient use in operational environments. This course is designed as a professional guide to teach professionals how to prepare basic language models for specific and customized tasks. Basic language models usually require targeted post-training processes to perform specialized tasks optimally and accurately. However, adapting open-source language models for specific applications in different domains often presents major challenges, such as graphics card memory limitations, high processing costs, and the complexity of post-training workflows. This course teaches how to overcome these obstacles and optimize language models for organizations. During this course, learners will first be introduced to the fundamental concepts of post-training, the tool ecosystem, and methods for preparing datasets, including generating dummy data. Next, they will learn how to implement fine-tuning with optimal memory management using LoRA and QLoRA methods, as well as preference alignment techniques such as DPO and GRPO. Finally, participants will learn how to quantize their tuned models and perform a comparative assessment of model accuracy and propagation speed. By the end of this course, individuals will have the knowledge and skills to deliver customized and optimized AI models.

What you will learn

• Preparing the dataset and using dummy data generation techniques for the post-training process
• Implementing fine-tuning methods with optimal memory usage such as LoRA and QLoRA
• Applying preference alignment techniques including DPO and GRPO methods to improve model performance
• Quantizing small language models and measuring accuracy against speed of dissemination and execution
• Deploying customized, small, open-source language models into operational environments

This course is suitable for people who:

• AI and machine learning engineers who want to customize language models
• Developers facing graphics card memory limitations and high processing costs
• Data scientists who want to build small language models for specialized applications in their organization
• People who want to learn compression, quantization, and optimization skills for open source models

Course details: Fine-Tuning and Optimizing Small Language Models

• Publisher: Pluralsight
• Instructor: Ned Bellavance
• Training level: Advanced
• Training duration: 3 hours and 28 minutes

Course topics

Course images

Sample course video

Download link

Download file – 781 MB

Rapidgator link

Download File – 781 MB

File(s) password: www.downloadly.ir

File size

781 MB