fix: catch TypeError when accessing think token properties#1141
Open
nightguarder wants to merge 2 commits into
Open
fix: catch TypeError when accessing think token properties#1141nightguarder wants to merge 2 commits into
nightguarder wants to merge 2 commits into
Conversation
…nking models This enables support for TranslateGemma and other non-thinking models whose tokenizer lacks _think_start_tokens. Previously blocked by TypeError, TranslateGemma-4b-it now works correctly via /v1/completions with a client-side chat template. Note: TranslateGemma uses a custom chat template requiring source_lang_code/target_lang_code fields that OMLX /v1/chat/completions does not support. The chat template must be applied client-side and the resulting prompt sent to /v1/completions instead. See the model's chat_template.jinja for the exact prompt format. Catch (ValueError, TypeError) on all three sites: - _detect_needs_think_prefix in scheduler step - ThinkingBudgetProcessor think_start_id resolution - _resolve_think_end_token_ids think_end_id resolution
cangming009
pushed a commit
to cangming009/omlx
that referenced
this pull request
May 10, 2026
mlx-lm#1171 changed the MiniMax M2 tool parser to return a list when a single <minimax:tool_call> block contains multiple <invoke>s. Without list/dict flattening in api/tool_calling.py, parallel tool calls were silently dropped via AttributeError swallowed by the existing except. Flatten parser results in the native path using the same isinstance(list) pattern already used in the Gemma 4 fallback. Add regression tests for single-dict, multi-list, and multi-block-multi-invoke cases. Also picks up BatchKVCache/BatchRotatingKVCache.extend() batch-dim fix (jundot#1141) and the tree_reduce import fix (jundot#1165) from the same bump.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fix TypeError for non_thinking models
Catch TypeError alongside ValueError in three spots in scheduler.py. Without this, /v1/completions crashes on TranslateGemma.
Explanation
Some models (e.g. TranslateGemma) don't have
_think_start_tokensinitialized, so accessing thethink_start_idproperty raises TypeError instead of ValueError.The three existing except ValueError blocks miss this, causing engine loop crashes when serving such models via
/v1/completions.Catch (ValueError, TypeError) on all three sites:
_detect_needs_think_prefixin scheduler step_resolve_think_end_token_ids think_end_idresolutionWhat this does
This enables support for TranslateGemma and other non-thinking models whose
tokenizer lacks
_think_start_tokens. Previously blocked byTypeError,Files modified
omlx/scheduler.py