Skip to content

docs(edge-agentic): add BFCL v4 quantization calibration dataset#424

Open
Palanivelg wants to merge 2 commits into
mlcommons:mainfrom
Palanivelg:edge-agentic-calibration-dataset
Open

docs(edge-agentic): add BFCL v4 quantization calibration dataset#424
Palanivelg wants to merge 2 commits into
mlcommons:mainfrom
Palanivelg:edge-agentic-calibration-dataset

Conversation

@Palanivelg

Copy link
Copy Markdown
Contributor

Adds examples/11_Edge_Agentic_Example/bfcl_calib.jsonl (364 chat/tool-aware BFCL v4 single-turn records) — the calibration set used for post-training quantization (NVFP4/FP8) of the Edge-Agentic reference model Qwen/Qwen3.6-27B. Calibration only; not scored and not part of the accuracy gate.

  • Documented in the example README under Quantization calibration dataset (schema, chat-template rendering, and the ~30% overlap with the ~995-sample gate).
  • Sits next to the perf dataset agentic_coding_2.5h.jsonl.

The MLPerf reference index (mlcommons/inference language/edge-agentic) will link this file; a companion PR there fixes the stale example link and adds the reference.

Add examples/11_Edge_Agentic_Example/bfcl_calib.jsonl (364 chat/tool-aware
BFCL v4 single-turn records) used only for post-training quantization
calibration -- not scored and not part of the accuracy gate -- and document
it in the example README.

Signed-off-by: Palanivelg <12385994+Palanivelg@users.noreply.github.com>
@Palanivelg
Palanivelg requested a review from a team July 17, 2026 22:57
@gemini-code-assist

Copy link
Copy Markdown
Contributor

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

@github-actions

Copy link
Copy Markdown

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

@codecov-commenter

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (main@598dbfd). Learn more about missing BASE report.

Additional details and impacted files
@@           Coverage Diff           @@
##             main     #424   +/-   ##
=======================================
  Coverage        ?   80.68%           
=======================================
  Files           ?      133           
  Lines           ?    17638           
  Branches        ?        0           
=======================================
  Hits            ?    14232           
  Misses          ?     3406           
  Partials        ?        0           

☔ 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.


---

## Quantization calibration dataset

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should the blog be updated to include the calibration dataset. We want to keep the two consistent to avoid any confusion.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants