Skip to content

fix(vlm): expose _tokenizer on VLMTokenizerWrapper for guided decoding #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.

Sign up for GitHub

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

Open
creedx12 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix(vlm): expose _tokenizer on VLMTokenizerWrapper for guided decoding #124

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions src/mlx_omni_server/chat/mlx/model_types.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -140,6 +140,12 @@ def __init__(self, processor):
else:
self.tokenizer = processor

# OutlinesLogitsProcessor accesses tokenizer._tokenizer
# (outlines_logits_processor.py), so expose it here too —
# otherwise json_schema/guided decoding with a VLM raises
# AttributeError: 'VLMTokenizerWrapper' has no '_tokenizer'.
self._tokenizer = self.tokenizer

# Load config from model path for apply_chat_template
# We'll get it from the model's config attribute if available
self.config = {"model_type": "gemma4"} # Default
Expand Down