Copy the template resolution logic from the base apply_chat_template to Voxtral#45117
Copy the template resolution logic from the base apply_chat_template to Voxtral#45117Rocketknight1 merged 3 commits intomainfrom
Conversation
fdb29a9 to
04a890a
Compare
|
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. |
zucchini-nlp
left a comment
There was a problem hiding this comment.
Great thanks a lot, caught up with model release and forgot about a failing test
c03a5ea to
e87bb2e
Compare
|
[For maintainers] Suggested jobs to run (before merge) run-slow: voxtral |
|
@zucchini-nlp my bad, I actually misread the Voxtral processor. Almost all of the actual handling is done inside Updated the code to just handle that instead! Can you take another look and I'll merge if you're happy? |
|
Hmm I guess voxtral has no special Those jinja-kwargs can't be passed over to processor, it'll return |
7312a5f to
b5bd46a
Compare
|
run-slow: voxtral |
|
This comment contains models: ["models/voxtral"] |
CI ResultsCommit Info
Model CI Report❌ 4 new failed tests from this PR 😭
|
|
Those tests are all failing already on main, so I don't think they should block this! We might want to check them in a future PR |
…to Voxtral (huggingface#45117) * Copy the template resolution logic from the base apply_chat_template to the voxtral override * Just return an empty set from the template variables scan instead * Revert other changes
Fixes #45084. We need to resolve the chat template in the Voxtral code to avoid
Nonebeing passed to_get_template_variables()!cc @zucchini-nlp, follow-up to #44881