Fix sparsity-only export emitting empty hf_quant_config.json

[kaix-nv]

What does this PR do?

Type of change: ?

Bug fix. Fix sparsity-only export writing hf_quant_config.json with null quant_algo.

Testing

Before your PR is "Ready for review"

Make sure you read and follow Contributor guidelines and your commits are signed (git commit -s -S).

Make sure you read and follow the Security Best Practices (e.g. avoiding hardcoded trust_remote_code=True, torch.load(..., weights_only=False), pickle, etc.).

• Is this change backward compatible?: ✅ / ❌ / N/A
• 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?: ✅ / ❌ / N/A
• Did you update Changelog?: ✅ / ❌ / N/A

Additional Information

Summary by CodeRabbit

• Bug Fixes

  • Export now correctly detects when a checkpoint is truly quantized and avoids writing unnecessary quantization metadata/files when no quantization algorithm is specified.

• Style

  • Reorganized imports across example modules for consistency and clearer import ordering.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 06f95962-f83f-49d4-8be3-e61f31b9ff2d

📥 Commits

Reviewing files that changed from the base of the PR and between e0901f7 and d306766.

📒 Files selected for processing (5)

• examples/vllm_serve/fakequant_worker.py
• examples/vllm_serve/vllm_ptq_utils.py
• examples/vllm_serve/vllm_reload_utils.py
• examples/vllm_serve/vllm_serve_fakequant.py
• modelopt/torch/export/unified_export_hf.py

✅ Files skipped from review due to trivial changes (4)

• examples/vllm_serve/vllm_ptq_utils.py
• examples/vllm_serve/vllm_reload_utils.py
• examples/vllm_serve/vllm_serve_fakequant.py
• examples/vllm_serve/fakequant_worker.py

---

📝 Walkthrough

Walkthrough

Multiple import statements were reordered across example files. Additionally, quantization export logic was changed to treat a checkpoint as quantized only when metadata includes quant_algo or kv_cache_quant_algo, avoiding hf_quant_config.json and quantization_config in config.json when neither is present.

Changes

vLLM Examples Import Reordering

Layer / File(s)	Summary
Import ordering examples/vllm_serve/fakequant_worker.py, examples/vllm_serve/vllm_ptq_utils.py, examples/vllm_serve/vllm_reload_utils.py, examples/vllm_serve/vllm_serve_fakequant.py	Reordered module imports within each file; no changes to runtime logic, public APIs, or behavior.

Quantization Export Logic

Layer / File(s)	Summary
Quantization detection modelopt/torch/export/unified_export_hf.py	Compute quantization_details = (hf_quant_config or {}).get("quantization", {}) and determine is_quantized_export by checking quant_algo or kv_cache_quant_algo presence.
Behavioral change modelopt/torch/export/unified_export_hf.py	If no quant algos present, set hf_quant_config = None to prevent writing hf_quant_config.json and avoid adding quantization_config to config.json; otherwise retain prior behavior and convert/write HF quant config.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

🚥 Pre-merge checks | ✅ 6

✅ Passed checks (6 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly describes the main bug fix in the changeset - preventing sparsity-only exports from creating empty hf_quant_config.json files. This directly corresponds to the primary logic change in unified_export_hf.py.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Security Anti-Patterns	✅ Passed	No security anti-patterns detected in modified files: torch.load, numpy.load, trust_remote_code, eval/exec, or nosec comments.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch kaix/sa_export_fix

Tip

💬 Introducing Slack Agent: The best way for teams to turn conversations into code.

Slack Agent is built on CodeRabbit's deep understanding of your code, so your team can collaborate across the entire SDLC without losing context.

• Generate code and open pull requests
• Plan features and break down work
• Investigate incidents and troubleshoot customer tickets together
• Automate recurring tasks and respond to alerts with triggers
• Summarize progress and report instantly

Built for teams:

• Shared memory across your entire org—no repeating context
• Per-thread sandboxes to safely plan and execute work
• Governance built-in—scoped access, auditability, and budget controls

One agent for your entire SDLC. Right inside Slack.

👉 Get started

---

_{Review rate limit: 9/10 reviews remaining, refill in 6 minutes.}

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

[github-actions]

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-05-04 21:11 UTC

[coderabbitai]

🧹 Nitpick comments (1)

modelopt/torch/export/unified_export_hf.py (1)

1192-1210: ⚡ Quick win

Please add a regression test for both export branches.

Add one test for sparsity-only export (assert no hf_quant_config.json / no quantization_config) and one for true quantized export (assert both are present).

As per coding guidelines: “Testing: ... add/adjust tests if the PR touches export output correctness.”

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@modelopt/torch/export/unified_export_hf.py` around lines 1192 - 1210, Add two
regression tests covering both branches in unified_export_hf.py: one for
sparsity-only export and one for true quantized export. For sparsity-only, call
the export function (the code path computing is_quantized_export using
quantization_details from hf_quant_config) with an hf_quant_config where
"quantization": {"quant_algo": None, "kv_cache_quant_algo": None} and assert
that no hf_quant_config.json file is written in export_dir and that the exported
metadata/returned hf_quant_config does not include a quantization_config. For
the quantized case, provide an hf_quant_config where "quantization":
{"quant_algo": "<some_algo>"} and assert hf_quant_config.json exists in
export_dir and that the export invoked convert_hf_quant_config_format (and the
exported metadata contains the expected quantization_config). Name tests to
reflect branches (e.g., test_export_sparsity_only_no_quant_config and
test_export_true_quantized_writes_quant_config) and clean up temporary
export_dir after assertions.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@modelopt/torch/export/unified_export_hf.py`:
- Around line 1192-1210: Add two regression tests covering both branches in
unified_export_hf.py: one for sparsity-only export and one for true quantized
export. For sparsity-only, call the export function (the code path computing
is_quantized_export using quantization_details from hf_quant_config) with an
hf_quant_config where "quantization": {"quant_algo": None,
"kv_cache_quant_algo": None} and assert that no hf_quant_config.json file is
written in export_dir and that the exported metadata/returned hf_quant_config
does not include a quantization_config. For the quantized case, provide an
hf_quant_config where "quantization": {"quant_algo": "<some_algo>"} and assert
hf_quant_config.json exists in export_dir and that the export invoked
convert_hf_quant_config_format (and the exported metadata contains the expected
quantization_config). Name tests to reflect branches (e.g.,
test_export_sparsity_only_no_quant_config and
test_export_true_quantized_writes_quant_config) and clean up temporary
export_dir after assertions.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 0fbcaa6b-538a-4d83-b54f-b53ae8af9034

📥 Commits

Reviewing files that changed from the base of the PR and between c07ac21 and e0901f7.

📒 Files selected for processing (5)

• examples/vllm_serve/fakequant_worker.py
• examples/vllm_serve/vllm_ptq_utils.py
• examples/vllm_serve/vllm_reload_utils.py
• examples/vllm_serve/vllm_serve_fakequant.py
• modelopt/torch/export/unified_export_hf.py

[codecov]

Codecov Report

❌ Patch coverage is 75.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 76.90%. Comparing base (acfab41) to head (09ffe48).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
modelopt/torch/export/unified_export_hf.py	75.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1375      +/-   ##
==========================================
- Coverage   76.90%   76.90%   -0.01%     
==========================================
  Files         471      471              
  Lines       50562    50565       +3     
==========================================
  Hits        38886    38886              
- Misses      11676    11679       +3     

Flag	Coverage Δ
examples	41.56% <75.00%> (+0.89%)	⬆️
gpu	59.67% <75.00%> (-0.66%)	⬇️
regression	14.89% <0.00%> (+0.06%)	⬆️
unit	52.80% <0.00%> (-0.01%)	⬇️

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

☔ View full report in Codecov by Sentry.
📢 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.

[shengliangxu]

LGTM, by the way do you know what codebase/system is still reading hf_quant_config.json?