Align README and repository header with trtcheck v1.1.0 - #24
Merged
Conversation
- opening no longer says trtcheck reports whether a model 'will convert'; it now states what static analysis can and cannot prove, with the verified-requires-runtime limitation up front - quick start leads with the two commands that matter (static, then --verify-runtime); JSON demoted to a secondary example - four-verdict table moved near the top with the four honesty rules - fixer section rewritten to v1.1.0 semantics (use-aware INT64, provable-inference Dropout, transactional execution, --fix before/after reporting, no implied runtime verification) - runtime evidence condensed into one conservative smoke-validation section (TensorRT 10.3.0, 7 fixtures, 5 builds / 2 expected failures, 7/7 agreement) with the dynamic-profile observation, linking to the full report - comparison table row no longer claims to 'predict the build outcome' - roadmap pruned of completed items - action examples pinned to @v1.1.0 (v1 floating tag still points at 1.0.0); example-consumer workflow updated - docs/index.md landing copy de-overclaimed; pyproject description refreshed (no release made for it)
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Why
Parts of the public copy predated the v1.1.0 verdict model and could
overstate what static analysis proves: the README opening said trtcheck
"reports whether the model will convert", the docs landing page promised
to tell you "whether the model will convert cleanly", and the comparison
table claimed it "predicts the build outcome". v1.1.0's whole point is
that static analysis returns
blocked/unverified/likelyand onlya real trtexec build returns
verified— the front page now says exactlythat.
What changed
limitation stated up front; four-verdict table plus its honesty rules
moved near the top.
trtcheck model.onnxand--verify-runtime;JSON demoted to a secondary example.
schema-aware fixers,
--fixbefore/after reporting, schema-2.0 ruleids + confidence +
verify_required+ plugin fallback ids.10.3.0, seven bounded fixtures, five builds, two expected parser
failures, 7/7 wrapper/direct agreement) including the TRT-10.3
degenerate auto-profile observation, linked to the full report.
@v1.1.0(the floatingv1tag still points at 1.0.0);docs/index.mdand thepyproject.tomldescription de-overclaimed;GitHub repository description updated.
Verification
Every README command was run against the public
trtcheck==1.1.0wheel in a fresh venv (flags checked against
--help; blocked exit code,JSON schema 2.0,
--fix, missing-trtexec path). Full suite 456passed / 1 skipped; mypy --strict, Black, isort clean; build +
twine check+ fresh-wheel smoke pass.No product behavior changed and no package version, tag, or release was
created.