LLM Basic - MOC

Input

• Tokenization
• Tokenizer in Modern LLMs
• BPE Encoding Complexity and Optimization
• Input Embedding in Modern Transformers
• LLM Hyperparameter
• LLM Precision About

Architecture

• Transformer in LLM
• Activated Params in MoE Models
• DeepSeek V4 Architecture Tricks
• Rotary Positional Embedding - Detail Explanation
• Casual Self Attention
• Multi-Head Attention
• Self-Attention Detail Example

Fine-tuning and Inference

• LoRA - Low Rand Adaptation
• KV Cache
• LLM cache hit

Evaluation

• perplexity

Agent and Appendix

• agent SKILL mechanism
• Embed vs. Transform

Parent MOC

• Large Language Model - MOC