Repository navigation
docs(metric-benchmark): locate results validation in the stage list - #151
Open
mstaats-nvidia wants to merge 1 commit into
Open
mstaats-nvidia wants to merge 1 commit into
mstaats-nvidia wants to merge 1 commit into
Conversation
The stage table lists the four numbered stages but gives no path from there to validate_results.py, which is documented further down under stage 1. Note the tool beneath the table and say why it is not numbered, so readers scanning the table can find it without implying a fifth stage. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Signed-off-by: mstaats <mstaats@nvidia.com>
asutermo
approved these changes
Oct 28, 2026
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds a four-line note beneath the stage table in
labs/metric-benchmark/docs/stages.mdpointingat
validate_results.py, and says why it is not one of the numbered stages.Why
stages.mdalready documents the tool, under stage 1's Results-directory contract subsection.But the summary table at the top lists only the four numbered stages, so a reader scanning the
table has no path to the one tool in the lab that exists today.
AGENTS.mdasks for theuser-facing page to cover a new tool a user must act on.
A numbered fifth row was the obvious fix and is the wrong one: the page's intro says "the four
executable stages below... one execution stage and three analysis stages", stage 1's prose says
validation "is not an additional scoring stage", and inserting a row would mean renumbering four
## N.headings. A note keeps the table's framing intact and makes the tool discoverable.Notes
feat/metric-benchmark-results(PR feat(metric-benchmark): specify and validate experiment results #105), sincevalidate_results.pyandresults-format.mdonly exist there. Retarget tomainonce feat(metric-benchmark): specify and validate experiment results #105 lands.tools.md#validate-a-results-directory(tools.md:140) and#results-directory-contract(stages.md:65).Checks
uv run --locked ruff check .— passeduv run --locked ruff format --check .— passed (305 files)make check-copyright-headers— passeduv run --locked pytest -q -rs -k metric_benchmark— 76 passedmake commit-check— DCO passedpytest -q -rswas not run to completion here (exceeded the 9-minute budget in thisenvironment);
make check-licensesneedsosv-scanner, which is not on PATH here. Both areCI's to confirm — this change touches one markdown paragraph.
🤖 Generated with Claude Code