Skip to content

armgit / laya

Public Imported from GitHub
0

Non-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ languages, with a router that picks the right checkpoint per request.

Merge pull request #1061 from keemsisi/feat/java-predict-tournament

Nandakishor69bd3cebbd85
diff --cc .github/workflows/java.yml
index 00c747d7,98c9a5f6..43fb8b9b
--- a/.github/workflows/java.yml
+++ b/.github/workflows/java.yml
@@@ -175,6 -175,6 +175,11 @@@ jobs
#
# This floor and the jdk-matrix one differ by one ON PURPOSE -- see that lane for why a
# shared number let a whole five-test class be deleted there.
++ # 1007/1008 is the pre-#1061 measurement, kept on purpose. #1057, #1059, #1060 and
++ # #1061 were each branched off main independently, so every one measured its own
++ # count and none measured the merged tree, which carries all four. The real figure
++ # is higher. The lower verified number keeps the floor binding and the lane green;
++ # an unmeasured figure in a gate is how a gate goes red for the wrong reason.
- name: Model-free suite was not vacuous
if: ${{ !cancelled() }}
run: >
@@@ -459,10 -459,11 +464,15 @@@
- name: predict
checkpoints: multilingual
graph-model: multilingual
- families: --only predict.json
+ families: --only predict.json --only tournament_golden.json
tests: >-
--tests com.convaiinnovations.laya.PredictParityTest
+ --tests com.convaiinnovations.laya.OneOptionGraphTest
+ --tests com.convaiinnovations.laya.TournamentParityTest
- # Measured locally: 9 PredictParityTest + 3 TournamentParityTest. Exact.
++ # 12 is a verified LOWER bound, not the union: #1059 and #1061 were branched
++ # off main independently and each measured 12 for its own pair, so neither
++ # measured all three classes together. Two of them use @TestFactory, so the
++ # executed count is not derivable statically. Re-measure on main and raise it.
min-tests: 12
timeout: 75
# Weight-bearing, and a SEPARATE cell rather than a second family in the one above.