vllm

Weights and KV are the two numbers everyone knows. The answer that gets the offer accounts for the other few gigabytes, says which ones vLLM measures and which it guesses, and knows where an OOM comes from when the sum looks fine.

A prefix cache only hits from the first token forward, so evicting the wrong block can strand gigabytes that stay resident and can never be hit. The answer derives how much, and which eviction order guarantees it never happens.

Per token, moving KV over PCIe is several times cheaper than recomputing it, yet vLLM recomputes. The answer derives the break-even bandwidth, shows how small copies fall below it, and names what makes recompute nearly free.

Chunked prefill keeps the decoders running, so the damage looks bounded. The walkthrough shows the inter-token gap climbing from 10 ms to a full second, new arrivals locked out for 8.6 seconds, and a slowdown that outlives the prefill.

The default of 1,024 sequences is never reached by a 2K-context workload on an 8B model, and is past the compute ridge for a 256-token one. The answer finds the three ceilings and picks the lowest.

Scheduling step n+1 before step n's tokens exist sounds impossible. The answer shows the placeholder trick that makes it work, what it costs at an end-of-sequence token, and the configurations where vLLM switches it off for you.

Each draft token is a real token in the step's budget, and verification turns a bandwidth-bound step compute-bound sooner. The answer shows the accounting, the rollback that makes rejection cheap, and the batch size where speculation starts losing.

Three knobs, one distribution, and an ordering question that most candidates have never thought about. The answer works one six-token example through every knob and shows the order changing which tokens can be sampled.

A drafter that costs nothing and needs no training sounds too good, and on open chat it does nothing. The answer traces a code-rename request step by step, counts accepted tokens under three match-length settings, and says when to turn it off.

Compute acceptance from counter snapshots, reject reset-contaminated windows, and handle EOS before inferring request-step counts. Then decide what a matched-load baseline must measure.

The model never sees the grammar; the sampler does, as a 16 KB bitmask per request per step. The answer builds that mask from an automaton, prices the naive version at ten milliseconds, and shows how precompilation removes almost all of it.

Four beams over a 1,000-token prompt cost 600 MiB if each keeps its own cache and as little as 156 MiB if they share blocks. The answer implements the fork, the copy-on-write and the prune, then shows what vLLM actually does today.

Five processes, one scheduler with no prefill phase, and a busy loop that should never wait on Python. The answer maps each stage to the component that owns it in the current source, not the 2023 paper.

Three flags, three different budgets, and one startup check that fails a 70B deployment on two H100s with a message most people misread. The answer walks the memory profile in the order vLLM runs it.

A graph freezes shapes, so engines capture dozens of them and round every batch up. The answer prices that rounding in the regime where decode lives and finds the real costs somewhere else.

A block's identity is a hash of its tokens and of every block before it. The answer traces one request through hashing, lookup, the last-token rule and eviction, with a runnable model of the lookup.

Both reuse KV for shared prefixes, and on a 2K-token chat prompt their hit rates differ by under one percent. The answer finds the differences that do matter, from eviction policy to whether a cache key means anything outside one process.

Decode each token on its own and an emoji streams as two replacement characters. The answer builds the offset-window decoder that fixes it, runs it, and adds the stop-string holdback that most streaming bugs come from.

Eight tenants, eight adapters, one base model and one batch. The answer explains why the extra cost scales with distinct adapters rather than requests, and which default serializes them.

The mask is a 16 KB bitmap per request. Whether it costs nothing or stalls every step depends on which thread computes it and what it is allowed to overlap with.

Same command line, new version, lower throughput. The answer is a method: capture the resolved configuration on both sides, diff the handful of defaults that move every quarter, then pin and bisect.

One engine is built to run a model anywhere with one user; the other is built to share a datacenter GPU across hundreds. The answer turns that into memory-bandwidth arithmetic and finds the cases where the laptop wins.

The scheduler did not change; the code that turns its decisions into GPU tensors did. The answer covers where per-request state now lives, why Triton became a hard dependency, and the feature list that silently puts you back on the old runner.

Short prompts, enormous replies, and a KV footprint that grows every step after admission. The answer sizes the batch from the pool, prices a preemption, and shows why per-user speed, not throughput, is the number to defend.

141 GB of weights on 160 GB of GPUs, and the engine refuses to start. The answer walks the five phases of engine startup, shows which one failed from the log, and does the arithmetic that says the model never fit the way it was configured.

The KV pool is preallocated, so the growth is somewhere the engine does not budget. The answer splits memory into what is fixed and what can grow, turns a growth rate into bytes per request, and uses that one number to rule out half the suspects.