Large Language Models

SpiralTrain BV

The course Large Language Models from SpiralTrain provides a comprehensive understanding of large language models (LLMs), from foundational architectures like transformers to advanced topics like fine

Price

from €1,699

Level

Not specified

Kind

Training

Subject

Automation and IT, Training and education

More information and registration at SpiralTrain BV

Description

Intro to LLMs This module introduces LLMs and their evolution, from GPT to BERT and T5. It explains transformers, attention, and tokenization. Participants explore training objectives, scaling laws, and key differences between pretraining and fine-tuning. Open source vs proprietary models are also compared. Model Architectures Participants learn LLM types: decoders, encoder-decoders, and key models like GPT, LLaMA, and PaLM. It covers training pipelines, optimizers, and precision formats. Tools like Hugging Face and Deepspeed are introduced, plus tuning techniques like LoRA and in-context learning. Training LLMs This module focuses on preparing and fine-tuning data for LLMs. It includes tokenizer setup, adapters, SFT, and avoiding overfitting. Participants learn metrics for evaluation, model alignment, and benchmarking. Hugging Face tools and best practices are emphasized. LLM Deployment Participants learn how to serve and optimize LLMs for production. Topics include quantization, distillation, and cloud deployment (AWS, Azure, GCP). Also covered are LangChain integration, embeddings, caching, and reducing inference costs with scalable strategies.