OpenFluke
w2a
w2a is our honesty check. It's the testing harness that measures Welvet and publishes the results, so you never have to just trust a marketing claim. If something only half-works, w2a says so.
Why a separate proof tool?
It's easy for software to claim it's fast and correct. w2a exists to actually check. It runs Welvet across every combination of settings, times how it performs, and stamps each result as fully working, partly working, or not yet done. Keeping this outside the engine means the tests can be brutally honest, and the engine itself stays clean and small.
Measured, not guessed
Timed runs across every numeric type, quant, and backend: real numbers, not vibes.
Honesty stamps
Each result is labelled: done, partial, or missing. No half-finished path gets counted as complete.
Public logs
Suite reports ship as downloadable release files, versioned to match the engine's scorecard.
The whole board, green
One end-to-end run on a desktop with an NVIDIA GTX 1650 SUPER. Every timed cell OK, every case passing, zero gaps and zero fails.
| Layer | Cells | OK | Cases |
|---|---|---|---|
| step | 32,458 | 32,458 | 11 |
| tween | 32,457 | 32,457 | 10 |
| dna | 16,159 | 16,159 | 6 |
| evolution | 16,152 | 16,152 | 6 |
| dense · mha · softmax · cnn · rnn · lstm · … | 2,802–2,847 ea | all OK | 14–19 ea |
| Total | 137,039 | 137,039 | 326 |
Run it yourself
cd welvet/w2a go run . # interactive suites go test ./tests/dense -v
Suite logs (suite.txt / suite.pdf) ship as GitHub Release assets, versioned to the
Welvet scorecard, not committed into git. The engine packages never contain tests; the proofs live here.