r8: IPA auxiliary-task experiment (phonological-layer claim, controlled) - #57
Open
ronaldtse wants to merge 4 commits into
Open
r8: IPA auxiliary-task experiment (phonological-layer claim, controlled)#57ronaldtse wants to merge 4 commits into
ronaldtse wants to merge 4 commits into
Conversation
…lled attached runs; fire-and-forget removes the local dependency
…sumed run completed 13,986 news units + 400 gold-2014 lines
ronaldtse
force-pushed
the
exp/r8-ipa-aux-clean
branch
from
August 27, 2026 02:22
0683edb to
3ace31b
Compare
Contributor
Author
|
CI note: all six failing jobs fail identically on the base branch itself ( |
…al); r7 entrypoint spawn r8 differs from r6 in exactly one variable (aux stream renders the same r5-units as broad-phonemic IPA instead of morphology). Result on the full 1,200-paragraph windowed zero-skip harness: r8 2.6588/1.5783 vs r6 2.5793/1.5317 vs r5 2.6775/1.5965. IPA-stream probe CER 0.0230 (EM 62/200) proves the second projection was learned — the comparison is not confounded by a failed aux task. Phonemic supervision is not the active ingredient; lexical-morphological knowledge is. r6 stays the teacher. r7 (news-domain) launched init from run-006-morph; entrypoint switched to spawn() (the r8 disconnect lesson). 19,716 steps ETA ~15.7h.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Replaces #56 (accidentally targeted main while carrying sprint-branch commits; that one is closed).
arabic_to_ipa.py— deterministic diacritized-Arabic → broad-phonemic IPA converter (rule-based; sun-letter assimilation, shadda gemination, long vowels, tanwin carriers, dagger-alaf lengthening; approximations documented). Verified on canonical phrases (ʔasːalaːmu ʕalajkum, kitaːbun mufiːdun, marħaban, ʔaʃːamsu tˤaːliʕatun, …), 20K units/strain_arabic_r8.py— the phonological-layer claim made controlled: r6's two-format multitask recipe with ONLY the auxiliary content changed (qalsadi morph tags → IPA of the same r5-units text, seeded ~25% aux share, init from r5, batch 2/accum 15, 1 epoch, A100-80GB). Same harness, hyperparams, and init as r6 — isolates the phonemic representation as the single variabler8-ipa-aux, app ap-ONuWKBrB7zfHUerMCM8Rx7)CI note: based on
sota-sprint-arabic(where the experiment lives) rather thanmain—main's lint/build jobs are red on the legacypython/arabic/python/hebrewtrees independent of this change; the sprint branch carries the green tree. Rebase to main together with the sprint branch when it merges.Test plan