Skip to content

Close the metadata tail: 100% units and references, +94 unit checks (N5)#15

Merged
Cuuper22 merged 1 commit into
mainfrom
claude/eager-cerf-kvmqta-n5
Jun 11, 2026
Merged

Close the metadata tail: 100% units and references, +94 unit checks (N5)#15
Cuuper22 merged 1 commit into
mainfrom
claude/eager-cerf-kvmqta-n5

Conversation

@Cuuper22

Copy link
Copy Markdown
Owner

N5 of 10: Close the metadata/provenance tail

Metadata coverage pass across 33 non-lithography scope modules (memory, cluster, economics, training, kernel, architecture, interconnect, precision, optimizer, parallelism, collective, gpu, thermal, noise). Lithography scopes were reserved for the N2 branch.

Observed coverage, before → after

Metric Before After
Variables with sp_units 1428 1493 (100%)
Variables with references 1324 1493 (100%)
Equations with references 878 959 (100%)
Equations with unit checks 799 893

References are real public documents (vendor datasheets, JEDEC specs, IEEE 754, standard texts); no fabricated provenance. Unit checks enabled only where the dimensional check passes.

Coordinator reconciliation on top of the interrupted agent's work

  • Registry snapshot test updated to the new coverage truth.
  • Curated unchecked-equation ledgers shrunk (cluster set now empty; kernel matmul/attention FLOP equations now checked; opex checked set gains run_power_cost and water_cost_rate).
  • Reverted check_units on the three optimizer schedule ordering inequalities: existing tests deliberately assert those feasibility relations stay unit-check-free, and overriding design tests is beyond a metadata pass's mandate.

Verification

  • Full pytest: 670 passed (metadata-only change; no new tests expected)
  • Audit gate: PASS
  • Live stats coverage block quoted above from this branch

Note: agent was interrupted by a session limit mid-reconciliation; coordinator finished the four bookkeeping tests, validated, committed, and pushed.

https://claude.ai/code/session_01Eu2JVnPFgMQftwYTP3cGQZ


Generated by Claude Code

Metadata coverage across 33 non-lithography scope modules (memory,
cluster, economics, training, kernel, architecture, interconnect,
precision, optimizer, parallelism, collective, gpu, thermal, noise):

Observed coverage before -> after:
  with_sp_units              1428 -> 1493 (every non-constant variable)
  with_references            1324 -> 1493 (every non-constant variable)
  equations_with_references   878 -> 959  (every equation)
  equations_with_unit_check   799 -> 893

References are real public documents (vendor datasheets, JEDEC specs,
IEEE 754, standard texts); no fabricated provenance. Unit checks were
enabled only where the dimensional check passes.

Reconciliation applied on top of the interrupted agent's work:
- Registry snapshot test updated to the new coverage truth.
- Curated unchecked-equation ledgers shrunk (cluster set now empty;
  kernel matmul/attention FLOP equations now checked; opex checked
  set gains run_power_cost and water_cost_rate).
- Reverted check_units on the three optimizer schedule ordering
  inequalities: existing tests deliberately assert those feasibility
  relations stay unit-check-free, and overriding design tests is
  beyond a metadata pass.

Full pytest: 670 passed. Audit gate: PASS.

https://claude.ai/code/session_01Eu2JVnPFgMQftwYTP3cGQZ
@Cuuper22 Cuuper22 merged commit f4e88f9 into main Jun 11, 2026
0 of 6 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants