chore(ci): bump codecov/codecov-action from 4 to 6 - #2
Open
dependabot[bot] wants to merge 1 commit into
Open
Conversation
Bumps [codecov/codecov-action](https://github.com/codecov/codecov-action) from 4 to 6. - [Release notes](https://github.com/codecov/codecov-action/releases) - [Changelog](https://github.com/codecov/codecov-action/blob/main/CHANGELOG.md) - [Commits](codecov/codecov-action@v4...v6) --- updated-dependencies: - dependency-name: codecov/codecov-action dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <[email protected]>
Author
LabelsThe following labels could not be found: Please fix the above issues or remove invalid values from |
abailey81
added a commit
that referenced
this pull request
Apr 27, 2026
…fy two-number framing, add re-runnable verifier
The phase-21 honesty audit caught me quoting the project's
*conservative-stress-test* perplexity (1725) as if it were the
headline number, when the *training-time response-only* perplexity
is **~ 147** (from `best_eval_loss = 4.987` in the checkpoint blob).
Both numbers are real and measure different things; only the second
should be the demo-cheat-sheet headline.
DOCUMENT FIXES
- docs/huawei/PRESENTER_CHEAT_SHEET.md ("What this prototype is NOT"
section): replaced the bare "perplexity ~ 1725" claim with the
two-number framing — 147 as the apples-to-apples comparable to
published SLMs, 1725 as the stress-test on broader distribution
with history-token loss included.
- docs/huawei/open_problems.md (#2 — full-corpus SLM retrain):
documented BOTH numbers in the background, and split the acceptance
criteria into "training-time held-out below 80" and
"stress-test below 600". Previously claimed only 1725 → 600,
which made the success bar artificially harsh.
- reports/slm_v2_eval.md: added a "Read this before quoting any
number" preamble at the top explaining the two metrics with a
one-line "use this number when…" guide. The stress-test 1725 is
now annotated inline ("← stress-test number, not the headline").
Added a "Headline number" subsection at the bottom citing
best_eval_loss = 4.987 and exp(4.987) ≈ 146.6 with the source.
Architecture line updated to disclose the 204.4 M unique vs
229.4 M state_dict spread (the difference is the tied-weight
LM-head + a few biases the model's `.num_parameters` dedupes).
NEW TOOLING — scripts/verify_numbers.py
- A re-runnable audit that loads each artefact on disk and asserts
every recruiter-facing claim matches. Caches no data; just
reads the live checkpoint, the live ONNX, the live KG, the live
corpus. Prints PASS / FAIL per claim, exits non-zero on drift.
- Current run: 22 / 22 claims verify. Specifically:
SLM v2 architecture (d_model 768, 12L, 12H, vocab 32 k)
SLM v2 step (18 000), eval_loss (4.987), ppl (~ 147)
SLM v2 params (~ 204 M unique, 229 M state_dict)
Qwen LoRA (rank 16, alpha 32, DoRA, 4 545/252, 1 704 steps × 3 epochs,
best_val_loss 5.36e-06, wall 9 656 s)
Encoder ONNX (FP32 441 KB, INT8 162 KB, 63 % reduction, parity
MAE 0.0006)
KG subjects (31), corpus pairs (974 k)
- Run anytime with: .venv/Scripts/python.exe scripts/verify_numbers.py
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
abailey81
added a commit
that referenced
this pull request
Apr 29, 2026
Iter 58 — the user reported the dashboard's 'Typing rhythm' tile still reading 0 ms after iter 41 + restart, even though composition cadence (1.29 s avg) was being recorded correctly on the same turn. Root cause: the iter 41 fix correctly routed JS-format ``iki_ms`` into the keystroke_buffer, but the buffer's first sampled keystroke event always has ``iki_ms = 0`` (no preceding keystroke). When a short message produced exactly ONE keystroke event (every 3rd keystroke is sampled — so a 3-key message lands on event #3 with keyTimings.length=2 → iki_ms is the gap between #2 and #3, OK; but when the very first sampled event is the only one, the buffer has [0]). The pre-fix server passed ``[0]`` straight to ``Pipeline._iki_stats``, which filtered the zero out and returned mean=0 — even though composition_metrics.keystroke_timings had real data right there. Three-level fallback: 1. server-side keystroke_buffer (filter to non-zero entries) 2. composition_metrics.keystroke_timings array 3. composition_metrics.mean_iki scalar (last resort: synthesize a single timing from the JS-precomputed mean) The dashboard's 'Typing rhythm' tile now never reads 0 ms when the JS client has any meaningful inter-key data — even on edge-case messages where the per-event sampler captured only zero-IKI entries.
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.
Bumps codecov/codecov-action from 4 to 6.
Release notes
Sourced from codecov/codecov-action's releases.
... (truncated)
Changelog
Sourced from codecov/codecov-action's changelog.
... (truncated)
Commits
57e3a13Th/6.0.0 (#1928)f67d33dRevert "Revert "build(deps): bump actions/github-script from 7.0.1 to 8.0.0""...75cd116chore(release): 5.5.4 (#1927)87d39f4Revert "build(deps): bump actions/github-script from 7.0.1 to 8.0.0" (#1926)1af5884chore(release): bump to 5.5.3 (#1922)c143300build(deps): bump actions/github-script from 7.0.1 to 8.0.0 (#1874)671740achore(release): 5.5.2 (#1902)96b38e9chore:disable_searchalignment (#1881)9b6d1f8check gpg only when skip-validation = false (#1894)5a10915chore(release): 5.5.1 (#1873)Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting
@dependabot rebase.Dependabot commands and options
You can trigger Dependabot actions by commenting on this PR:
@dependabot rebasewill rebase this PR@dependabot recreatewill recreate this PR, overwriting any edits that have been made to it@dependabot show <dependency name> ignore conditionswill show all of the ignore conditions of the specified dependency@dependabot ignore this major versionwill close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this minor versionwill close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this dependencywill close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)