[Minor] fix: do not requantize the scales in FP8 scale sweep calibration

[Fridah-nv]

What does this PR do?

Type of change: ?

Overview: ?

Usage

# Add a code snippet demonstrating how to use this

Testing

Before your PR is "Ready for review"

• Make sure you read and follow Contributor guidelines and your commits are signed.
• Is this change backward compatible?: Yes/No
• Did you write any new necessary tests?: Yes/No
• Did you add or update any necessary documentation?: Yes/No
• Did you update Changelog?: Yes/No

Additional Information

Summary by CodeRabbit

Release Notes

• New Features

  • Added conditional pre-quantization optimization for static block quantization when scale sweep is enabled.
  • Enhanced calibration flow with configurable scale quantization behavior based on optimization settings.

• Improvements

  • Updated quantization calibration documentation to clarify scale behavior and optimization interactions.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[coderabbitai]

Important

Review skipped

Auto incremental reviews are disabled on this repository.

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.

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

• 🔍 Trigger a full review

📝 Walkthrough

Walkthrough

These changes implement an optimized FP8 quantization path for NVFP4 static per-block quantization. When fp8_scale_sweep is enabled, the quantizer skips dynamic FP8 scale quantization by setting a flag that propagates through the calibrator and into the fake quantization path, using pre-computed scales instead.

Changes

Cohort / File(s)	Summary
FP8 scale sweep configuration modelopt/torch/quantization/model_calib.py	Added conditional logic to enable pre-quantization optimization for NVFP4 static per-block quantization when fp8_scale_sweep is enabled; sets skip_fp8_scale_quant flag on the quantizer before MSE calibrator construction. Updated docstring to clarify FP8 scale value count (128 → 126 valid values).
Fake quantization flag handling modelopt/torch/quantization/nn/modules/tensor_quantizer.py	Introduced skip_scale_quant local variable derived from block_sizes["skip_fp8_scale_quant"] in the 2-bit static block quantization path; passes this flag to static_blockwise_fp4_fake_quant instead of always using False.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Title check	✅ Passed	The title accurately describes the main change: preventing scale requantization in FP8 scale sweep calibration, which is the core objective evident in both modified files.

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

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 73.82%. Comparing base (81b67dd) to head (1caa24f).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #825   +/-   ##
=======================================
  Coverage   73.82%   73.82%           
=======================================
  Files         193      193           
  Lines       19745    19745           
=======================================
  Hits        14577    14577           
  Misses       5168     5168           

☔ 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.

[realAsma]

@Fridah-nv can you please add [Minor] tag in the PR title?

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.