[Example]: Dspark W4A16 PTQ script

[h-guo18]

What does this PR do?

Type of change: new example

Adds examples/speculative_decoding/scripts/quantize_drafter.py, a CLI that quantizes an exported speculative-decoding drafter to W4A16 NVFP4.

Needs no calibration data (the preset's per-block scales are dynamic, so mtq.quantize is called without a forward_loop) and no modeling code — exported drafters such as nvidia/MiniMax-M3-DSpark have no importable model class, so each 2-D weight is wrapped in a throwaway nn.Linear under its checkpoint key and ModelOpt's usual quantizer_name patterns select over those names. Works for any drafter layout. Config comes from QUANT_CFG_CHOICES; packing and scales go through the existing export helpers. No library code touched.

Usage

python scripts/quantize_drafter.py \
    --drafter_path nvidia/MiniMax-M3-DSpark \
    --export_path ./MiniMax-M3-DSpark-W4A16

w4a16_nvfp4: 9.98 GiB -> 3.51 GiB

Writes a unified-HF checkpoint (w.weight / w.weight_scale / w.weight_scale_2) plus hf_quant_config.json. lm_head is excluded by default — unlike on a base model it is 37% of this drafter's parameters, so --quantize_lm_head is a real lever (~1.9 GiB), but measure AL first.

Testing

No automated test yet (see checklist). Validated on nvidia/MiniMax-M3-DSpark: runs in 67 s on CPU, 9.98 GiB (fp32) -> 3.51 GiB; all 43 quantized tensors round-trip within 0.0952 relative error; the 29 untouched tensors are bit-identical to bf16(source); the 86 added keys are exclusively scale buffers. pre-commit passes.

Before your PR is "Ready for review"

• Is this change backward compatible?: ✅
• If you copied code from any other sources or added a new PIP dependency, did you follow guidance in CONTRIBUTING.md: N/A
• Did you write any new necessary tests?: ❌ — validated manually as above. Can add a tests/examples/speculative_decoding/ test over a small synthetic drafter if wanted before merge.
• Did you update Changelog?: N/A (example-only)
• Did you get Claude approval on this PR?: ❌ (not yet run)

Additional Information

Whether vLLM's DSpark drafter loader honours quantization_config on the draft model is not verified here.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 2ad43cef-3d80-4196-9c48-5156f96f2b66

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 66.83%. Comparing base (87c9f8c) to head (2ec20d0).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2027   +/-   ##
=======================================
  Coverage   66.83%   66.83%           
=======================================
  Files         519      519           
  Lines       58916    58916           
=======================================
  Hits        39376    39376           
  Misses      19540    19540           

Flag	Coverage Δ
unit	54.89% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.