Skip to content

Port gHashTag/trios:crates/trios-train-cpu/src/bin/main.rs (Rust, 5 functions) to specs/port/trios/crates/trios-train-cp - #9258

Merged
gHashTag merged 2 commits into
masterfrom
queen-9235
Oct 12, 2026
Merged

gHashTag merged 2 commits into
masterfrom
queen-9235

Conversation

@gHashTag

Copy link
Copy Markdown
Owner

Closes #9235

Written by a bee on queen-9235 and published by tools/queen/publish.py. The branch itself is the bee's; the second commit is the coordination entry every pull request must add, which a bee has no way to know about.

1 file changed, 640 insertions(+)

🤖 Generated with Claude Code

Trinity Bee and others added 2 commits October 12, 2026 11:52
Re-authors gHashTag/trios crates/trios-train-cpu/src/bin/main.rs
(196 lines at 8b229e9489ee) as
specs/port/trios/crates/trios-train-cpu/src/bin/main.t27, one function
per original function: find_arg, parse_config, load_data,
run_training_loop, main.

Decisions are ported as pure functions, each under test: prefix
scanning and number parsing (find_arg specialized per value type -
.t27 has no generics; the f64 variant keeps the name), parse_config
defaults and flag precedence (--encoder-lr= over --lr=, --no-jepa,
--optimizer=muon), the six assert! lines as config_is_valid, the
fallback corpus and %128 token map and empty-check of load_data, and
the 90/10 split, validation-set choice, window offset, warmup step,
report cadence and best-bpb update of run_training_loop. main's
Result-to-exit-code decision is main_exit_code.

load_data, run_training_loop and main keep their names with undefined
bodies (file read, the unported training stack, argv/banner/clock
plumbing); no test calls them, and the generated Zig carries no stub
markers.

Deviations, named: trial_id/agent_id label strings become given-flags
(.t27 cannot slice a string; main.rs 88-89); f32 widens to f64;
float parsing covers the CLI's decimal shapes, not Rust's full
from_str grammar; seq_offset defines 0 where the original would panic
on modulo zero when dl <= SEQ + 1.

Closes #9235
A pull request must add exactly one docs/now entry and a bee has no way
to know that: its brief names a boundary file and acceptance criteria,
and docs/now/ is neither. The publisher adds it rather than failing the
gate.

Closes #9235

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@github-actions

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-10-12 12:17:12 UTC

Summary

Status Count
Total Open PRs 50
PRs with Failing Checks 47
PRs with All Checks Green 3
READY 0
FAILING 47
PENDING 0
NO CHECKS YET 0

These columns do not partition: 0 + 47 + 0 + 0 = 47, and there are 50 open PRs. A PR is being counted twice or not at all.

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=e769ef021713 != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

@tri-claw tri-claw Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

Reviewer bee verdict for head 7a6ddf5b3f64a640862ffaad96abeed675426c96 (tools/bees/reviewer.py, zai glm-4.7-flash, 7 turns, 137 s; then glm-4.5-flash, 4 turns, 105 s).

BEE-VERDICT: APPROVE
summary: All 6 acceptance criteria met, no new defects, boundary respected
criterion: 1. test -f specs/port/trios/crates/trios-train-cpu/src/bin/main.t27 && echo present prints present -- met -- evidence: brief.md:894
criterion: 2. grep -cE '^\s*(pub )?fn (find_arg|parse_config|load_data|run_training_loop|main)\(' specs/port/trios/crates/trios-train-cpu/src/bin/main.t27 prints 5 -- met -- evidence: brief.md:895
criterion: 3. t27c gen ... prints 0 "not yet implemented" and at least 13 lines -- met -- evidence: brief.md:896-897
criterion: 4. t27c spec-status does not print NOPARSE -- met -- evidence: brief.md:898
criterion: 5. grep -cE 'test ...' prints at least 5 -- met -- evidence: brief.md:899
criterion: 6. t27c test-report prints 0 BLOCKED -- met -- evidence: brief.md:900
discounted-check: Corpus ratchet (expected-failure ledger) -- was already red on master for the same reason (brief.md:530)
discounted-check: Documented t27c subcommands exist -- was already red on master (brief.md:582)
discounted-check: cli-tri -- not run on master for last 6 commits (brief.md:790)
discounted-check: coverage -- was already red on master (brief.md:686)
discounted-check: duplicate-bodies -- was already red on master (brief.md:634)
discounted-check: spec-guards -- was already red on master (brief.md:738)
discounted-check: t27b-native-ratchet -- was already red on master (brief.md:842)

Evidence

Perfect. I have confirmed:

  1. All 6 acceptance criteria are MET (verified by runner facts in brief.md lines 894-900)
  2. No new defects introduced by this head - all 7 red checks were already red on master
  3. Boundary respected - only 1 spec file changed (the docs/now file is publisher-generated)

Second, independent review (glm-4.5-flash): APPROVE -- Successfully ported 5 functions from Rust to .t27 with all acceptance criteria met

@tri-claw tri-claw Bot added the bee-reviewed A reviewer bee reviewed and verified this PR at its current head; the only merge signal (#5525) label Oct 12, 2026
@gHashTag
gHashTag enabled auto-merge (squash) October 12, 2026 14:25
@gHashTag
gHashTag merged commit 9d4424b into master Oct 12, 2026
25 of 32 checks passed
gHashTag added a commit that referenced this pull request Oct 12, 2026
…0 names, 21 read and judged (11 DRIFT, 10 DISTINCT) (#9293)

Port wave (#9258 trios-train-cpu, #9259 trios-crypto, #9188 hir-cost-models) grew
the conflicted-type set by 21; observed stable at exactly 21 across two master
tips (21c6171, f235d90; no port landed after 16:20:49Z). Every NEW name was
read at both definition sites, evidence quoting declarations and import edges,
per the #9145 standard -- then blessed in the same PR as the verdicts (a bless
alone leaves the classified gate red).

- docs/reports/type_conflicts.json: ledger 89 -> 110
- docs/reports/type_conflicts_classified.json: +21 verdicts (11 DRIFT,
  10 DISTINCT), census 62/48, ASCII-escaped
- docs/TYPE_CONFLICTS.md: +21 table rows, section counts 51->62 / 38->48

Verified on the branch: tri types ratchet RATCHET: CLEAN; tri types classified
OK (every conflicted name has a written verdict, no stale verdicts).

Refs #9281

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
gHashTag added a commit that referenced this pull request Oct 12, 2026
…83 (Refs #9281) (#9303)

The W628 expected-failure ledger went over cap (77 > 75) and showed six
UNEXPECTED FAILURES on master (run 38290342182, identical on dbf1920 and
90fc25b -- standing, not caused by the #9299 bless).

Six ported specs, evidence from the run's own ratchet.log artifact:

- five W469 refusals, [N][]const u8 locals the ports must keep faithful to
  their sources: depin_rest #7014 (fields, line 60), tri_asm_base #9211
  (lines, 136), sweep #7070 (values, 139), trios-server main #9254 (types,
  325), trios-train-cpu main #9258 (args, 467)
- gen_lru_cache_impl #8923: ported identifiers lower to a bare Verilog
  keyword `end` (verilog-no-keyword-decl, generated line 779)

The over-cap state was left by #7689/#7795/#9176, which hand-ledgered three
W469 entries without raising the cap while the job was red at an earlier
step; the single raise 75 -> 83 covers both, one reviewable event.

Refs #9281.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

bee-reviewed A reviewer bee reviewed and verified this PR at its current head; the only merge signal (#5525)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Port gHashTag/trios:crates/trios-train-cpu/src/bin/main.rs (Rust, 5 functions) to specs/port/trios/crates/trios-train-cpu/src/bin/main.t27

2 participants