Add int8_sq back to auto_quant support list

[cjluo-nv]

What does this PR do?

minor change

Overview: ?

Add int8_sq back to auto_quant support list. We will just export the final checkpoint as the tensorrt_llm checkpoint.

Summary by CodeRabbit

• New Features
  • Added support for INT8 SmoothQuant in the auto-quantization workflow, enabling selection of the int8_sq format for model compression. This expands quantization options and can improve performance and memory efficiency on compatible hardware. No changes to public APIs or user-facing workflows; existing configurations continue to work as before, with the new format available as an additional choice.

[coderabbitai]

Walkthrough

Adds "int8_sq" to the qformat_list in auto_quantize within examples/llm_ptq/hf_ptq.py, enabling the int8_smoothquant configuration to be selected through the auto-quantize flow. No other logic, control flow, or API signatures changed.

Changes

Cohort / File(s)	Summary
Quantization option update examples/llm_ptq/hf_ptq.py	Include "int8_sq" in qformat_list within auto_quantize to allow selecting INT8_SMOOTHQUANT_CFG via the auto-quantize path.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~2 minutes

Poem

A nibble of bits, a hop to queue,
I toggled a switch to let INT8 through.
SmoothQuant whispers, “I’m in the mix!”
The carrots compile, no extra tricks.
With one small string, we bound and flew—
Byte-sized dreams in int8 hue. 🥕🐇

Pre-merge checks and finishing touches

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	You can run @coderabbitai generate docstrings to improve docstring coverage.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title Check	✅ Passed	The title "Add int8_sq back to auto_quant support list" is a short, single sentence that directly summarizes the primary change—restoring int8_sq to the auto_quant support list—so it accurately reflects the changeset and is clear for a teammate scanning history.

✨ Finishing touches

• 📝 Generate Docstrings

🧪 Generate unit tests

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch cjluo-nv-patch-1

---

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.}

[coderabbitai]

Actionable comments posted: 0

🧹 Nitpick comments (2)

examples/llm_ptq/hf_ptq.py (2)

92-106: int8_sq re‑added to auto-quant support: LGTM; tighten wording and minor var rename

Change looks good. Two tiny cleanups to reduce confusion and shadowing:

• Clarify the assertion message (this path may export TRT‑LLM checkpoints when int8_sq is used).
• Avoid shadowing the parameter name in the comprehension.

-    assert all(
-        qformat
+    assert all(
+        fmt
         in [
             "fp8",
             "int8_sq",
             "int4_awq",
             "nvfp4",
             "nvfp4_awq",
             "w4a8_awq",
             "fp8_pb_wo",
             "w4a8_mxfp4_fp8",
             "nvfp4_mlp_only",
         ]
-        for qformat in qformat_list
-    ), "One or more quantization formats provided are not supported for unified checkpoint export"
+        for fmt in qformat_list
+    ), "One or more quantization formats provided are not supported by the auto-quantize export path"

Is exclusion of "fp8_pc_pt" from this allow-list (while present in QUANT_CFG_CHOICES) intentional for auto-quant? If yes, consider a brief comment above to prevent future regressions.

---

120-121: Avoid shadowing the built-in name format in list comprehension

Minor readability nit: don’t shadow Python’s built-in format().

-        quantization_formats=[QUANT_CFG_CHOICES[format] for format in qformat_list],
+        quantization_formats=[QUANT_CFG_CHOICES[fmt] for fmt in qformat_list],

📜 Review details

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 4c36abe and e54ce4e.

📒 Files selected for processing (1)

• examples/llm_ptq/hf_ptq.py (1 hunks)

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (4)

• GitHub Check: linux
• GitHub Check: wait-checks / wait
• GitHub Check: build-docs
• GitHub Check: code-quality

🔇 Additional comments (1)

examples/llm_ptq/hf_ptq.py (1)

588-596: Export behavior consistent with PR description

Condition explicitly routes int8_sq to TensorRT‑LLM checkpoint export. Matches the PR statement about final checkpoint format. No action needed.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 73.83%. Comparing base (4c36abe) to head (e54ce4e).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #345   +/-   ##
=======================================
  Coverage   73.83%   73.83%           
=======================================
  Files         172      172           
  Lines       17453    17453           
=======================================
  Hits        12887    12887           
  Misses       4566     4566           

☔ 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.