NXP backend: Documentation for QAT

[StrycekSimon]

Summary

Adds documentation NeutronQuantizer support for QAT.
Adds train support for QAT mode in calibrate_and_quantize function.

Test plan

Mostly documentation related changes - no tests added.
The quantization related function is partially covered by already existing tests.

cc @robert-kalmar @JakeStevens @digantdesai

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18228

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 2 New Failures, 1 Pending, 2 Unrelated Failures

As of commit a04917e with merge base faa5903 ():

NEW FAILURES - The following jobs have failed:

• pull / unittest-arm-backend-with-no-deps (test_pytest_models_tosa) / linux-job (gh)
  RuntimeError: Command docker exec -t 09a572c264a0540b45ea0043aa8c4eccea60e52ff93f77191ba0b4f0c90925b7 /exec failed with exit code 1
• pull / unittest-editable / macos / macos-job (gh)
  backends/xnnpack/test/ops/test_conv2d.py::TestConv2d::test_fp16_conv2d

FLAKY - The following jobs failed but were likely due to flakiness present on trunk:

• pull / test-qnn-wheel-packages-linux (3.12) / linux-job (gh) (detected as infra flaky with no log or failing log classifier)
• pull / unittest-wasm-bindings (--enable-etdump) / linux-job (gh) (detected as infra flaky with no log or failing log classifier)

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[StrycekSimon]

Since we mention the calibrate_and_quantize helper function as our public API in our documentation page, I added training support in this function to correspond with our new QAT feature. Can it be part of this PR?

[StrycekSimon]

@robert-kalmar @MartinPavella @roman-janik-nxp can some of you please take a look at this PR?

[Copilot]

Pull request overview

Adds NXP (NeutronQuantizer) Quantization Aware Training (QAT) documentation and updates the NXP quantization helper to optionally run a training callback when is_qat=True.

Changes:

• Link NXP quantization docs from the quantization overview page.
• Add a new “Quantization Aware Training” section to NXP quantization documentation (workflow + limitations).
• Extend calibrate_and_quantize to support an optional QAT training callback and optional calibration_inputs.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 7 comments.

File	Description
docs/source/quantization-overview.md	Adds QAT/PTQ mention and links to NXP quantization docs.
docs/source/backends/nxp/nxp-quantization.md	Adds QAT documentation, examples, and known limitation/workaround notes.
backends/nxp/quantizer/utils.py	Adds optional QAT training via train_callback and makes calibration_inputs optional.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[robert-kalmar]

Since we mention the calibrate_and_quantize helper function as our public API in our documentation page, I added training support in this function to correspond with our new QAT feature. Can it be part of this PR?

I don't prefer that, as it introduce a new feature in the pipeline and the objective is, this PR to get cherry-picked to the release branch where new features are not strongly not preferred - #17016 (comment).

Move the "training support" in separate PR to land only on main branch.

The documentation is valid even without that change, if you remove the "Alternatively" part - https://github.com/pytorch/executorch/pull/18228/changes#diff-a510398a70db14eec60cc59d9f317799bc7aede898e9042c6e3ba00f4cac10cdR198

[StrycekSimon]

Okay, I will remove it. That's why I asked.

[robert-kalmar]

LGTM.

[robert-kalmar]

@pytorchbot cherry-pick --onto release/1.2 -c docs

[pytorchbot]

Cherry picking #18228

The cherry pick PR is at #18282 The following tracker issues are updated:

• [v1.2.0] Release Schedule and Tracker #17016 (comment)

Details for Dev Infra team

Raised by workflow job