Skip to content

[FR] Fix grammar and translation errors in chapter2/5 - #1295

Open
IROFI wants to merge 5 commits into
huggingface:mainfrom
IROFI:fix-fr-chapter2-5-translation
Open

[FR] Fix grammar and translation errors in chapter2/5#1295
IROFI wants to merge 5 commits into
huggingface:mainfrom
IROFI:fix-fr-chapter2-5-translation

Conversation

@IROFI

@IROFI IROFI commented Aug 25, 2026

Copy link
Copy Markdown

Six line fixes in chapters/fr/chapter2/5.mdx. No line added or removed, no code snippet touched.

The intro question list

line before after why
37 comment gérer de plusieurs séquences ? comment gérer plusieurs séquences ? gérer takes a direct object, the partitive de is ungrammatical
38 comment gérer de plusieurs séquences de longueurs différentes ? comment gérer plusieurs séquences de longueurs différentes ? same
40 existe-t-il une séquence trop longue ? une séquence peut-elle être trop longue ? see below

The third bullet ("les indices du vocabulaire sont-ils les seules entrées…") is a faithful translation and is left as is.

On line 40: the English is idiomatic — "Is there such a thing as too long a sequence?" asks whether a sequence can be too long. The literal French reads as "does one particular over-long sequence exist?", which is not the question the "Séquences plus longues" section goes on to answer. Worth noting that es, it and pt made the same literal choice, so a French reviewer may well prefer to keep it — happy to drop this one commit if so.

Three other errors in the same file

line before after why
340 Les transformers acceptent en entrée que des séquences… Les transformers n'acceptent en entrée que des séquences… the ne of the restrictive ne … que was missing; dropping it is common in speech but incorrect in writing
97 appliqué à une `séquence` appliqué à une `sequence` the backticked identifier is the sequence variable defined in the snippet above, so it must not be translated
95 Pourquoi cela a échoué ? Pourquoi cela a-t-il échoué ? missing subject inversion in a direct question

Verification

  • Built chapters/fr with doc-builder build --html — builds clean, and the rendered list and snippet were checked against the page.
  • make quality reports the same result as main (pre-existing failures, from a newer black than the one pinned in CI).
  • Emphasis, backticks, links and MDX tags are unchanged on every edited line.

IROFI added 5 commits August 25, 2026 11:41
`gérer` takes a direct object, so the partitive `de` is ungrammatical here.
Matches the English source ("How do we handle multiple sequences?") and the
section title, which already says "Manipulation de plusieurs séquences".
"Les transformers acceptent en entrée que des séquences d'une longueur
limitée" drops the `ne` of the restrictive `ne ... que`. Omitting it is common
in spoken French but incorrect in writing, and it leaves the sentence reading
as a plain affirmative until the reader reaches `que`.

English source: "With Transformer models, there is a limit to the lengths of
the sequences we can pass the models."
The backticked identifier refers to the `sequence` variable defined in the
code snippet just above, so translating it to `séquence` breaks the link with
the code the sentence is describing.
…choué ?"

A direct question in written French needs the inversion; "Pourquoi cela a
échoué ?" belongs to the spoken register.
The English question is idiomatic: "Is there such a thing as too long a
sequence?" -- it asks whether a sequence *can* be too long. The literal
"existe-t-il une séquence trop longue ?" reads as "does one particular
over-long sequence exist?", which is not the question the section answers.

"Une séquence peut-elle être trop longue ?" keeps the meaning and matches the
"Séquences plus longues" section it points to.
@IROFI

IROFI commented Aug 25, 2026

Copy link
Copy Markdown
Author

@lewtun @stevhliu for review, and @lbourdois as the maintainer of the French translation — this one needs a native-speaker call, in particular on line 40 (existe-t-il une séquence trop longue ?). The two grammar fixes (lines 37-38 and 340) and the untranslated sequence identifier (line 97) should be uncontroversial; the line 40 rewording is a judgement call and I am happy to drop that commit if you prefer the literal form.

@HuggingFaceDocBuilderDev

Copy link
Copy Markdown

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants