Translate what was left of tier 1 - #50
Merged
Merged
Conversation
The 29 tier 1 entries that stopped being passthrough when the identifier rule narrowed were English headings, so they went into the queue as untranslated and this is the run that answered them. 27 calls over server2 and server3 in 0.3h at 91.7 calls per hour, 414 entries accepted, 26 refused and 6 out of rungs. Tier 1 untranslated goes 31 to 4. The 4 are the entries that ran out of rungs there, and A04 names them rather than leaving them silent. The same run cleared batches left queued from earlier sessions in tiers 2 to 6, which is why machine goes 737 to 867 across the corpus and not only in tier 1. L07 goes 687 to 816 because the fleet answered 129 more batches on gpt-5-6-mini, which is issue #37 on the translator repo and not a change here.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The 29 tier 1 entries that stopped being passthrough in #48 were English headings, so they went into the queue as untranslated. This is the run that answered them.
The run
27 calls over server2 and server3 in 0.3h at 91.7 calls per hour. 414 entries accepted, 26 refused, 6 out of rungs, 0 empty and 0 failed.
P05format specifiersP08fence or ruleP01protected spansP02placeholder markersThe ladder did its job: 21 of the 32 refusals were caught on the first rung and never reached the corpus.
Tier 1
The 4 are the entries that ran out of rungs.
A04names them with 4 findings rather than leaving them silent, which is what that check is for.Outside tier 1
The same run cleared batches left queued from earlier sessions, so machine goes 737 to 867 across the corpus rather than only in tier 1. Tier 3 machine goes 22 to 54, tier 5 41 to 97.
L07goes 687 to 816 because the fleet answered 129 more batches ongpt-5-6-mini. That is tamnd/python-docs-vi-translator#37 and not a change here.apply --checkis clean andS08is 0. Still nine hard checks failing, the same nine as before this run, so the audit step stays advisory.