-
Notifications
You must be signed in to change notification settings - Fork 124
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
resolve complicated chat templates during tokenizer saving #1151
resolve complicated chat templates during tokenizer saving #1151
Conversation
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. |
d4ba3f2
to
ccf3287
Compare
ccf3287
to
7cd9492
Compare
I am fine with the changes. @eaidova, can you please describe the idea briefly for the rest of the reviewers? |
Depends on huggingface/optimum-intel#1151 Close openvinotoolkit#1663 Ticket 161313
Depends on huggingface/optimum-intel#1151 Close openvinotoolkit#1663 Ticket 161313
Depends on huggingface/optimum-intel#1151 Close #1663 Ticket 161313
Depends on huggingface/optimum-intel#1151 Close #1663 Ticket 161313
Depends on huggingface/optimum-intel#1151 Close #1663 Ticket 161313
Depends on huggingface/optimum-intel#1151 Close #1663 Ticket 161313
Depends on huggingface/optimum-intel#1151 Close #1663 Ticket 161313
56823cd
to
df7e385
Compare
df7e385
to
bbc48e4
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM! Left a very minor comment
Co-authored-by: Ella Charlaix <80481427+echarlaix@users.noreply.github.com>
@echarlaix, thanks for review. I applied your comment. Could this PR be merged? |
What does this PR do?
Currently we have some issues with running some models with chat templates using C++ openvino API due to its chat template complexity parcing using C++. For smooth experience, we allow redefine them by own. To simplify such models usage, we want to provide C++ compatible simplified chat templates during tokenizer conversion (no impact on inference model with original python tokenizer) for several known and popular cases.
Additionally, it was found that sometimes running multimodal models (VLM), main chat template provided in processor instead of tokenizer as the result, our tokenizer conversion API ignore it. This PR fixes issue for such models.
Before submitting