quantized-export skill

Quantized Export

The last stop after checkpoint-promotion hands off a PROMOTE verdict: a checkpoint that cleared the four-stage gate still isn't deployed until it's exported in the right format for its target runtime and proven to still work post-export. A REJECT verdict never reaches this skill — export starts only from a promoted checkpoint.

Input: a promoted checkpoint (or LoRA adapter) plus the target deployment surface — GPU class, serving stack, and whether long-context/code/math workloads are in scope. Output format: an exported artifact in the chosen format plus a smoke-test diff report comparing 3–5 golden outputs pre-export and post-export.

Format Map

Pick format by hardware and deployment shape, not by habit — the wrong pick either wastes throughput headroom or breaks silently on specific workloads (see Workload Overrides).

and newer.** It preserves near-bf16 quality at roughly half the memory, and it's the safe first choice whenever the target GPU supports it and no edge-device constraint applies.

• **FP8 is the default on Hopper-class GPUs

FP8 hardware support. GPTQ is superseded for new deployments — don't reach for it on a fresh export; AWQ has better accuracy retention at the same bit width and wider current tooling support.

• AWQ INT4 targets older GPUs that predate

an imatrix, is the edge/llama.cpp format.** Use it for local or CPU-adjacent deployment, not for GPU-serving throughput — it optimizes for footprint, not tokens/sec on a datacenter GPU.

• GGUF with Q4K_M quantization, built from

deployments only — and explicitly NOT on GB10. NVFP4 on SM121 (GB10) runs ~32% slower than FP8 because the hardware lacks a native cvt.e2m1x2 path unless the kernel is compiled sm121a. Choosing NVFP4 on a GB10 target is a regression, not an upgrade — pick FP8 there instead.

• **NVFP4 is for Blackwell-at-scale

from quant format.** A merged export folds the adapter into the base weights: larger artifact, no base-model dependency at serve time. LoRA-only keeps the adapter separate: much smaller artifact, but the serving stack must load the exact same base model alongside it — a mismatched or wrong-revision base silently changes outputs. Pick merged when artifact portability matters more than storage; pick LoRA-only when disk footprint or multi-adapter serving matters more.

• **Merged vs. LoRA-only is a separate axis

Worked Picks

The core format-selection tradeoff, read as a lookup table for common scenarios:

# quick decision snippet — see the table above for the full map
hopper_or_newer: fp8
older_gpu: awq-int4
edge_llama_cpp: gguf-q4_k_m+imatrix
gb10_any_workload: fp8-vllm-nightly   # never nvfp4 on GB10

Workload Overrides

The Format Map above is a default, not a rule that survives every workload. Long-context, code, and math workloads break at INT4 — quantization error compounds across long sequences and precise token-level reasoning in ways that don't show up on short, generic prompts. For any of these three workload classes, stay on FP8 or W8A8 even if the target hardware would otherwise justify INT4 on cost grounds.

similar broad-knowledge benchmarks — they don't stress the failure mode. Measure with the actual task evals — the goldens and graders from eval-harness-first, run through the exported artifact — because INT4 degradation on long-context, code, or math shows up as task-specific failures (dropped context, broken syntax, arithmetic errors) well before it moves a knowledge benchmark.

• Don't validate this override with MMLU or

export on one of these three workload classes, the fix is switching format, not re-tuning the quantization recipe — AWQ and GPTQ variants at the same bit width share the same compounding-error failure mode on these workloads.

• If a task eval regresses after an INT4

The Smoke Test

Export bugs are silent at the file level — a malformed export still produces a loadable artifact, so file-existence checks prove nothing. The smoke test is mandatory for every export, with no exception for a format that "should just work":

target runtime** — vLLM for FP8/AWQ, llama.cpp for GGUF, not a quick sanity load in a different framework than the one that will serve it in production.

1. **Load the exported artifact in its actual

pull these from the same eval/goldens.jsonl eval-harness-first maintains, not a fresh ad hoc set.

1. Run 3–5 golden prompts through it —

pre-export generation for the same prompt, same deterministic sampling settings — greedy decoding (temperature 0) and a fixed seed, persisted and reused between the pre- and post-export runs, not just nominally identical config. For a lossless export, byte match is the gate — any diff is a bug. For a lossy** (quantized) export, byte match is expected to fail; the gate is task-grader verdict agreement instead — see references/export-commands.md's Smoke-Test Script Skeleton.

1. **Compare each output against the

Run this as a gate, not a manual check:

python smoke_test.py "$EXPORT_PATH" \
    eval/goldens.jsonl pre-export-outputs.jsonl
# non-zero exit on any pre/post mismatch

Failure Signatures

What export bugs actually look like, not a clean pass/fail flag:

run-on output — the chat template baked into the export doesn't match the one the checkpoint was trained and evaluated against, so turn boundaries or special tokens land in the wrong place.

• Template mismatch presents as garbled or

presents as off-template or semantically nonsensical output that still looks fluent — the output head lost precision it needed even though the rest of the network quantized cleanly.

• Wrong quantization applied to lmhead**

Never ship an export that skipped this step — a checkpoint's PROMOTE verdict says the un-exported checkpoint is good; it says nothing about the export pipeline. Re-run on any quant-method or runtime version bump, not only after the first export. Runnable command sequences for every format plus the smoke-test script skeleton: references/export-commands.md.

Related Skills

upstream source for this skill. A checkpoint without a PROMOTE verdict doesn't reach export.

• checkpoint-promotion — the only valid

eval/goldens.jsonl this skill's smoke test draws its 3–5 prompts from, and the task evals the Workload Overrides section requires for long-context/code/math validation.

• eval-harness-first — owns the

references/model-catalog.md is the place to check hardware-class assumptions (which GPU generations a base model targets) before picking a format off the Format Map above.

• finetuning-method-selection — its

Spark users: on GB10, GGUF via llama.cpp works well for local serving, and FP8 serving via vLLM nightly builds is the other proven path — NVFP4 is the one format to avoid there (see the Format Map exception above). Once the dgx-spark-ops plugin is installed, defer Spark-specific serving and thermal questions to its skills rather than re-deriving them here.