Inference & Serving · core
Speculative Decoding
Draft models, token verification, acceptance rates, tree speculation, latency, and quality preservation.
Mental model
A cheap draft proposes several tokens and the target model verifies them in parallel. Speedup depends on acceptance rate and verification cost while preserving the target distribution.
How to study Speculative Decoding
Begin by restating the mental model in your own words, then connect it to a concrete system you have built or operated. Name the mechanism, the constraint it addresses, and the trade-off it introduces. Use Fast Inference from Transformers via Speculative Decoding to check details, but close the source before writing your explanation. Retrieval is the learning step; rereading is only preparation.
Next, compare Speculative Decoding with FlashAttention & Attention Kernels, GPU Utilization. Ask what changes in correctness, latency, resource use, operability, and failure recovery. Complete Design exercise: Speculative Decoding and preserve the command, input, output, and one failed attempt as evidence. Finish by explaining the idea without jargon to someone who has not studied the track.
Proof of understanding
- Explain the mechanism from first principles and identify the state it reads or changes.
- Give one situation where the concept is the right choice and one where it is not.
- Predict a realistic failure mode before running the drill, then compare the prediction with evidence.
- Connect the result to a roadmap or build artifact instead of treating the concept as isolated trivia.
Learn from primary sources
Practice and explain it back
Design exercise: Speculative Decoding
Draft models, token verification, acceptance rates, tree speculation, latency, and quality preservation. Implement designOutline() returning non-empty values for: draftStrategy, verification, acceptanceRate. Each value must name a concrete mechanism or decision.
Expected evidence: A design outline with draftStrategy, verification, acceptanceRate plus an explicit failure mode or trade-off.
Open the interactive drill →Review prompts
- Speculative decoding runs two models yet is faster, and the output distribution is unchanged. Explain both claims.
Build evidence
Synthesize: Inference & Serving
Build a production mental model for inference engines, memory, kernels, routing, hardware, and serving economics. Produce one working system, benchmark, or evidence-backed design that integrates the path.
- Implements or precisely models the core mechanisms from all three milestones
- Includes at least one injected failure or adversarial case and demonstrates recovery
- Reports quality, latency, resource, reliability, or usability measurements relevant to the domain
- Ships a concise architecture note explaining decisions, trade-offs, and remaining risks