Thank you for sending your enquiry! One of our team members will contact you shortly.
Thank you for sending your booking! One of our team members will contact you shortly.
Course Outline
AI Sovereignty and Local LLM Deployment
- Risks associated with cloud LLMs: data retention policies, potential training on user inputs, and foreign jurisdiction implications.
- Ollama architecture overview: model server, registry functions, and OpenAI-compatible API support.
- Comparative analysis with vLLM, llama.cpp, and Text Generation Inference.
- Model licensing details for Llama, Mistral, Qwen, and Gemma.
Installation and Hardware Configuration
- Installing Ollama on Linux systems with CUDA and ROCm support.
- CPU-only fallback options and AVX/AVX2 optimization techniques.
- Docker deployment strategies and persistent volume mapping.
- Multi-GPU configuration and VRAM allocation best practices.
Model Management
- Fetching models from the Ollama registry (e.g., ollama pull llama3).
- Importing GGUF models sourced from HuggingFace and TheBloke.
- Evaluating quantization levels: tradeoffs between Q4_K_M, Q5_K_M, and Q8_0.
- Managing model switching and understanding concurrent model loading limitations.
Custom Modelfiles
- Syntax for writing Modelfiles: utilizing FROM, PARAMETER, SYSTEM, and TEMPLATE directives.
- Tuning hyperparameters such as temperature, top_p, and repeat_penalty.
- Engineering system prompts to define role-specific behaviors.
- Building and publishing custom models to the local registry.
API Integration
- Utilizing the OpenAI-compatible /v1/chat/completions endpoint.
- Implementing streaming responses and JSON mode.
- Integrating with frameworks like LangChain, LlamaIndex, and custom applications.
- Managing authentication and rate limiting via reverse proxy configurations.
Performance Optimization
- Sizing context windows and managing KV cache effectively.
- Executing batch inference and handling parallel requests.
- Allocating CPU threads and ensuring NUMA awareness.
- Monitoring GPU utilization and memory pressure levels.
Security and Compliance
- Establishing network isolation for model serving endpoints.
- Implementing input filtering and output moderation pipelines.
- Maintaining audit logs of prompts and generated completions.
- Verifying model provenance through hash verification.
Requirements
- Intermediate proficiency in Linux and container administration.
- High-level understanding of machine learning principles and transformer models.
- Familiarity with REST APIs and JSON data formats.
Target Audience
- AI engineers and developers looking to transition away from cloud LLM APIs.
- Organizations bound by data sensitivity regulations that prohibit the use of cloud models.
- Government and defense teams requiring air-gapped language model solutions.
14 Hours
