This lesson is part of the local ML Learning Lab collection.
OpenHands Core Concepts
Core concepts, architecture, and engineering challenges in OpenHands.
Open lessonCatalog
Lesson pages turn the source posts into structured study units with metadata, prerequisites, and review hooks. Use them directly, or follow them through a roadmap or course.
Official Collection
This lesson is part of the local ML Learning Lab collection.
Core concepts, architecture, and engineering challenges in OpenHands.
Open lessonThis lesson is part of the local ML Learning Lab collection.
CodeAct paper fundamentals, ReAct vs CodeAct framing, and design concepts behind OpenHands.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Startup lifecycle and run_controller initialization flow in OpenHands.
Open lessonThis lesson is part of the local ML Learning Lab collection.
OpenHands server architecture, Socket.IO flow, and session orchestration internals.
Open lessonThis lesson is part of the local ML Learning Lab collection.
OpenHands interaction and session internals: ConversationManager, WebSession, AgentSession, and oh_user_action flow.
Open lessonThis lesson is part of the local ML Learning Lab collection.
OpenHands EventStream internals: subscription, event distribution, Action/Observation flow, and AgentThinkAction.
Open lessonThis lesson is part of the local ML Learning Lab collection.
OpenHands function-calling internals: tool design, action mapping, and robust parsing of LLM tool calls.
Open lessonThis lesson is part of the local ML Learning Lab collection.
OpenHands Agent internals: state management, agent types, state lifecycle, and LLM adapter design.
Open lessonThis lesson is part of the local ML Learning Lab collection.
OpenHands CodeActAgent internals: design principles, tools, context engineering, and workflow.
Open lessonThis lesson is part of the local ML Learning Lab collection.
OpenHands AgentController internals: lifecycle control, routing, callbacks, observability, and robust execution control.
Open lessonThis lesson is part of the local ML Learning Lab collection.
OpenHands Runtime internals: sandbox execution, runtime types, event flow, and core code paths.
Open lessonThis lesson is part of the local ML Learning Lab collection.
OpenHands runtime deep dive: plugin system, execution system, and browser environment internals.
Open lessonThis lesson is part of the local ML Learning Lab collection.
OpenHands memory internals: layered memory architecture, View/ConversationMemory/Condenser workflow, and implementation details.
Open lessonThis lesson is part of the local ML Learning Lab collection.
OpenHands microagents deep dive: architecture, delegation workflow, trigger mechanisms, and implementation details.
Open lessonOfficial Collection
This lesson is part of the local ML Learning Lab collection.
A deep dive into attention mechanism foundations: seq2seq background, CNN/RNN limitations, attention principles, and historical evolution to Transformer.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Tokenization fundamentals in Transformers: vocabulary construction, tokenizers, BPE/WordPiece/Unigram, and newer token-free directions.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Transformer embedding fundamentals: from vectorization to trainable embeddings, implementation details, and modern text-embedding model evolution.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Transformer data processing pipeline: dataset choices, vocabulary/tokenizers, batch construction, masks, and training data loading in Harvard code.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Transformer overall architecture: workflow, attention modules, construction from Harvard code, and theoretical perspectives.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Transformer positional encoding: why it is needed, design evolution, sinusoidal encoding analysis, and NoPE debates.
Open lessonThis lesson is part of the local ML Learning Lab collection.
RoPE positional encoding, derivation, properties, extrapolation, and implementation.
Open lessonThis lesson is part of the local ML Learning Lab collection.
APE vs RPE in Transformers: differences, representative methods, and relative-position design patterns.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Self-attention in Transformers: principles, implementation details, scaling/softmax analysis, and modern optimization directions.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Transformer masks: padding mask, sequence/causal mask, implementation details, data flow, and advanced sample-packing strategies.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Multi-head self-attention in Transformers: motivation, principles, implementation details, and modern head-composition improvements.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Transformer encoder and decoder internals: architecture, data flow, cross-attention, and decoder-only design tradeoffs.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Transformer training and inference in practice: teacher forcing, masks, dropout, label smoothing, learning rate scheduling, and parallelism.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Feed-Forward Networks (FFN) in Transformers: structure, implementation, function, knowledge utilization, and optimization evolution.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Residual connections and normalization in Transformers: ResNet intuition, BatchNorm vs LayerNorm, Pre-Norm vs Post-Norm, implementations, and recent variants.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Transformer generator heads, softmax, decoding strategies, sampling parameters, logits analysis, and weight sharing.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Transformer parameter counts, memory usage, activations, FLOPs, KV cache, and optimization directions.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Autoregressive inference redundancy, KV cache, prefill vs decode, implementation, and resource usage.
Open lessonThis lesson is part of the local ML Learning Lab collection.
MQA and GQA: MHA review, shared KV heads, grouped-query attention, implementation details, memory and speed tradeoffs, conversion, and optimization variants.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Length extrapolation in Transformers and LLMs: position encoding methods, RoPE extrapolation, PI, NTK-aware interpolation, YaRN, and Giraffe.
Open lessonThis lesson is part of the local ML Learning Lab collection.
FlashAttention, online softmax, tiling, IO-awareness, and memory-efficient exact attention.
Open lessonThis lesson is part of the local ML Learning Lab collection.
FlashAttention V2, Flash-Decoding, Flash-Mask, and FlashAttention-3.
Open lessonThis lesson is part of the local ML Learning Lab collection.
KV Cache optimization: metrics, memory crisis, formula-based compression, stage-aware optimization, memory management, and scheduling.
Open lessonThis lesson is part of the local ML Learning Lab collection.
KV cache optimization for long text sequences: sparsification, token reuse, prefix reuse, retrieval-based schemes, and long-context KV management.
Open lessonThis lesson is part of the local ML Learning Lab collection.
KV cache optimization through PD separation or merging: static batching, ORCA, Sarathi, DistServe, SplitWise, MemServe, TetriInfer, and Mooncake.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Mixture-of-Experts (MoE): conditional computation, routing, experts, load balancing, implementation, and parallel inference.
Open lessonThis lesson is part of the local ML Learning Lab collection.
LoRA: PEFT, low-rank adaptation, rank, initialization, implementation, optimization, and continual learning.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Quantization fundamentals for Transformer LLMs: compression background, numerical representations, PTQ/QAT workflows, calibration, granularity, and acceleration.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Large model quantization fundamentals: outliers, superweights, massive activations, PTQ, QAT, and common quantization strategies.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Large model quantization schemes across 8-bit, 4-bit, and low-bit settings, including LLM.int8(), ZeroQuant, SmoothQuant, GPTQ, AWQ, LLM-QAT, QLoRA, FlatQuant, SqueezeLLM, SpQR, BitNet, and OneBit.
Open lessonThis lesson is part of the local ML Learning Lab collection.
DeepSeek MLA: low-rank KV compression, weight absorption, decoupled RoPE, resource tradeoffs, implementation details, and conversions from GQA and MHA.
Open lessonThis lesson is part of the local ML Learning Lab collection.
DeepSeek MoE: load balancing, fine-grained and shared experts, DeepSeek V1/V2/V3 routing, MoD, LoRA hybrids, and efficient fine-tuning.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Speculative decoding, speculative sampling, blockwise parallel decoding, token tree verification, and Hugging Face implementation details.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Medusa: multi-decoding heads, tree attention, typical acceptance, sparse tree construction, training strategies, and decoding flow.
Open lessonThis lesson is part of the local ML Learning Lab collection.
Lookahead decoding: Jacobi decoding, n-gram pool, 2D window, parallel verification, and llama.cpp implementation details.
Open lessonThis lesson is part of the local ML Learning Lab collection.
DeepSeek MTP: EAGLE, HASS, classical multi-token prediction, DeepSeek’s causal-chain design, formulas, and the vLLM implementation.
Open lesson