OpenNMT / OpenNMT/CTranslate2

The result for T5 is not the same

Open
#1,918 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
C++
Stars
4.7k
Forks
537
Avg merge
12h 12m
Merged PRs (30d)
4

Description

I'm typing here:
translate English to German: Together, they own about 14% of its publicly listed shares and control 56% of its stockholder voting power through super-voting stock.
And I get the result:
Zusammen besitzen sie etwa 14 % ihrer börsennotierten Aktien und kontrollieren 56 % ihrer Stimmrechtsbeteiligung durch Überabstimmungen.

Then I do the same with the converted model using this guide and get:
Gemeinsam besitzen sie etwa 14 % ihrer börsennotierten Aktien und kontrollieren 56 % ihrer Stimmrechte durch Überabstimmungen.

Why is the result not the same?

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reproducing the two outputs from the linked Hugging Face T5 page and the CTranslate2 Transformers conversion guide. Compare the model, conversion settings, and inference inputs used in each path; done means identifying the cause of the differing German translation or documenting the missing information needed to reproduce it.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.