Community-PR close comment: open a detailed issue, not a PR - #296
Merged
Conversation
…ot a PR The auto-close comment previously invited re-opening the change as a PR. Redirect non-optimization reporters to open a detailed issue instead (bug/impact/proposed fix), since the harness/tooling/docs are maintainer-owned. The training/dataset re-open note stays for anyone who actually meant an optimization submission. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Dexterity104
pushed a commit
to Dexterity104/SparkDistill
that referenced
this pull request
Jul 29, 2026
docs/research-summary.md (previously untracked, never committed) was stale relative to the last two weeks of merged work — commit it and bring it current: - Sec 4 (mining tracks): caveat that live entrius/gittensor weights still pay training and dataset tracks equally; the 2x-training PR is open, not merged (gittensor#1660). - Sec 7 (frontiers): replace the ancient "Hopper bucket never seeded" bug story with the current one — training PRs never auto-merged (fixed gittensor-model-hub#289) and the post-merge ledger write couldn't land on protected main (fixed gittensor-model-hub#290) combined to nearly cause a false REJECT on gittensor-model-hub#288, the first real Blackwell frontier candidate, when its 1-hour GPU attestation expired during the ~2h it sat waiting for a manual merge. - Sec 8 (TritonBench fairness): correct the "kernels are compiled and run, not judged by prose alone" overclaim — only 35% of the composite is execution-gated, and that 35% is self-graded (the model writes and grades its own torch.allclose check). Documents two concrete, still-open findings: Hopper is hard-rejected from execution entirely (bench_config.py require_blackwell_gpu, still present), and only 3 harness problems exist (fixed to be honest about it in gittensor-model-hub#282; populating more is roadmap gittensor-model-hub#227). - Sec 18 + Key PRs: root-cause the "frontiers low on exec/correctness" risk entry; refresh the PR list with what's actually landed recently. CONTRIBUTING.md had drifted from actual CI behavior: it still said non-quality PRs "we'll review and merge good ones" / harness PRs are "held for manual review" — true for maintainers, but community non-training/dataset PRs are now auto-closed by CI (gittensor-model-hub#295/gittensor-model-hub#296), not queued for review. The auto-close comment itself says "see CONTRIBUTING.md" for the policy, which didn't actually explain it. Add a "Found a bug or have an idea?" section pointing community contributors at issues instead, and correct the two paragraphs that implied review-then-maybe-merge. Also: a note at the top of the PR template (seen before submitting, not after), and an ISSUE_TEMPLATE/bug_report.md so the "open a detailed issue instead" redirect lands somewhere structured instead of a blank box — codifies the What/Reproduction/Impact/ Expected/Fix shape already used successfully in recent issues (gittensor-model-hub#204, gittensor-model-hub#212, gittensor-model-hub#214, gittensor-model-hub#237, gittensor-model-hub#241). Full suite: 384 passed (docs-only, no code changes). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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.
Follow-up to #295. The community-PR auto-close comment now directs non-optimization reporters to open a detailed issue (bug / impact / proposed fix) instead of a pull request — the harness/tooling/docs are maintainer-owned. The training/dataset re-open note stays for anyone who actually intended an optimization submission.
Comment content isn't asserted in tests (only that a comment is posted), so no test change is needed. Policy tests still pass; ruff clean.
Also applied the policy to the existing backlog: closed the 17 open community non-optimization PRs (#224, #229, #230, #238, #242, #246, #255–#263, #292–#294) with this comment; kept #277 (maintainer).