Local LLM
Local LLM: Weight memory estimate
Estimate parameter storage from parameter count and precision before accounting for overhead.
Read article : Local LLM: Weight memory estimateAI, Cyber Defense & IT.
Learn. Build. Defend.
Explore 2246 articles on AI, cybersecurity, networking and IT. 25 detailed practical guides now include concrete inputs, step-by-step procedures, expected results and troubleshooting. The remaining 2221 articles retain their existing content.
AI Cyber Defense · 200 advanced articles
Defend against AI-assisted attacks with practical guidance for cybersecurity and network professionals. Ten reading paths contain 20 articles each, with threat scenarios, evidence requirements, control designs and authorized validation plans. The collection includes protection for AI systems used by defenders and businesses. Behavioral evidence is kept separate from claims about an attacker's use of AI.
AI Agents & Local LLM · 300 techniques
Build reliable agents and run language models on your own infrastructure. Explore 160 AI Agents techniques and 140 Local LLM techniques across 15 reading paths, with 20 techniques per path. Each guide includes an implementation action, a validation exercise, platform context and official documentation for further reading.
Data monitoring · 200 practical articles
Diagnose stale reports, late logs, slow queries and unreliable alerts. Explore 40 Power BI articles, 40 Splunk articles, 40 SQL Server articles, plus 20 each on PostgreSQL, MySQL, pipeline freshness and data quality, and alerting and observability. Each guide connects a specific symptom to evidence, a focused correction and validation.
CCNA exam preparation · 200 techniques
Prepare for CCNA 200-301 v1.1 with 100 focused topics, each paired with a recall guide and an original practice exercise. Build understanding through packet tracing, subnet calculations, lab evidence, troubleshooting and spaced review. These are independent study guides; 200 is the collection size, not the number of questions on the exam.
Reviewed 2 November 2026. Check Cisco’s current exam information before booking. Reading paths are study groupings and do not represent official exam weights.
Advanced Network & Infrastructure · 200 articles
Ten focused reading paths with 20 articles each. Start from routing and fabric behavior, then explore transport, name resolution, security, host performance, storage, availability and incident evidence.
AI assistant user guides
Choose a platform to explore 50 guides, from a first task brief to reusable workflows. Microsoft Copilot guides cover the consumer and Microsoft 365 experiences; in-app features depend on your account and license.
Find a useful read
Matching articles (20)
Clear filtersResults for “reading-path:agents-local-11”
Local LLM
Local LLM: KV cache headroom
Reserve memory for context and active sequences in addition to model weights.
Read article : Local LLM: KV cache headroomLocal LLM
Local LLM: Context allocation fit
Choose context length from the task distribution and model support rather than the advertised maximum.
Read article : Local LLM: Context allocation fitLocal LLM
Local LLM: Quantization format support
Check that the runtime and hardware support the exact quantization format.
Read article : Local LLM: Quantization format supportLocal LLM
Local LLM: Precision quality gates
Define error thresholds for extraction, math and tool arguments before accepting lower precision.
Read article : Local LLM: Precision quality gatesLocal LLM
Local LLM: Calibration data relevance
Use representative non-sensitive inputs when a quantization method requires calibration.
Read article : Local LLM: Calibration data relevanceLocal LLM
Local LLM: CPU offload tradeoff
Measure the latency cost when some model computation or weights move to CPU.
Read article : Local LLM: CPU offload tradeoffLocal LLM
Local LLM: GPU memory fragmentation
Track allocation failures after repeated workload changes and compare a clean process.
Read article : Local LLM: GPU memory fragmentationLocal LLM
Local LLM: Memory-mapped model loading
Measure startup and resident memory behavior with the runtime's supported file loading mode.
Read article : Local LLM: Memory-mapped model loadingLocal LLM
Local LLM: Swap pressure detection
Track host swap activity and major page faults during inference.
Read article : Local LLM: Swap pressure detectionLocal LLM
Local LLM: Concurrent context footprint
Measure memory at the actual number of simultaneously active sequences.
Read article : Local LLM: Concurrent context footprintLocal LLM
Local LLM: Embedding and chat coexistence
Budget memory separately for embedding and generation services sharing a host.
Read article : Local LLM: Embedding and chat coexistenceStart with the foundations
Our original short guides to AI, account protection, and recovery.