Fix Kimi-VL Chat template #376
Merged
Conversation
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
There was a problem hiding this comment.
Pull Request Overview
Adds support for downloading .jinja template files by including the extension in the allowed pattern list for model artifacts.
- Added
"*.jinja"to the file pattern whitelist inget_model_path.
Comments suppressed due to low confidence (2)
mlx_vlm/utils.py:140
- Add or update a test case to verify that
.jinjafiles are correctly matched and downloaded byget_model_path.
"*.jinja",
mlx_vlm/utils.py:140
- The function's docstring or inline comment should be updated to mention that
.jinjafiles are now supported in the allowed patterns.
"*.jinja",
This was referenced May 27, 2025
Closed
breynolds007
added a commit
to breynolds007/mlx-vlm
that referenced
this pull request
Jun 22, 2025
fix: black and isort requests fix: Default to False for backward-compatible behaviour docs: Add docs for new feature FastAPI server (Blaizzy#321) * adding server (chat and batch endpoints not tested - generate OK with streaming on or off ) * cleanup * progress on openAi api * openai working * removing batch processing * cleaning openai model * removing adapter_path and resize_shape and seed --------- Co-authored-by: Prince Canuma <prince.gdt@gmail.com> feat: Resume training from saved adapter files (Blaizzy#327) * feat: Add ability to resume from a saved adapter path. * docs: Fix docs to use hyphen and not underscore * nit: Update parameter to use --adapter-path * fix: Order of imports --------- Co-authored-by: Prince Canuma <prince.gdt@gmail.com> docs: Fix command-line params in LORA.md to use dash separators (Blaizzy#331) Fix internVL3 (Blaizzy#333) fix models/internvl_chat/vision.py transpose issue (Blaizzy#340) [gemma3] Improve generation quality when using multiple images (Blaizzy#342) * working * expand image mask in mlx * fix batch size squeezing * batch size fix * working masked scatter 2d * 3d masked scatter * refactor masked_scatter * integrate mlx_masked_scatter into gemma3 * remove copy * fix whitespace * del masked_scatter.py * del bos workaround * Add better MLX masked scatter --------- Co-authored-by: Prince Canuma <prince.gdt@gmail.com> Update utils.py (Blaizzy#322) remove duplicate get_model_path (Blaizzy#343) [phi3_v] Remove logit dtype conversion from float32 to uint32 (Blaizzy#329) * [phi3_v] Remove logit dtype conversion from float32 to uint32 * Update mlx_vlm/models/phi3_v/phi3_v.py --------- Co-authored-by: Prince Canuma <prince.gdt@gmail.com> Make Gemma3 and Qwen2.5VL text-image input merging functions public static (Blaizzy#335) Co-authored-by: Prince Canuma <prince.gdt@gmail.com> fix: remove self reference in static method prepare_inputs_for_multimodal (Blaizzy#348) Fix VLMs upcasting (Blaizzy#350) * fix gemma upcasting * fix internvl * fix kimi_vl * fix tests bump version (Blaizzy#351) Fixes Qwen2 VL Position Id (Blaizzy#319) * handles case for position ids in qwen2 * remove duplicate * fix: generation command * resets rope_deltas * address comments * passes image and vdo grid to kwargs * refactor config and remove duplicate input_ids * fix tests --------- Co-authored-by: Prince Canuma <prince.gdt@gmail.com> Pixtral: make merge_input_ids function public static (Blaizzy#355) Add test for StoppingCriteria reset (Blaizzy#357) Fix typo in POST request example (Blaizzy#358) Fix early exit for StoppingCriteria (Blaizzy#356) Add max_position_embeddings to TextConfig (Blaizzy#360) Fix revision argument not passed in load (Blaizzy#361) * Forward revision argument in load function * format Add initial MkDocs setup (Blaizzy#362) * Link examples to GitHub * add GH actions * format Update deploy-docs.yml (Blaizzy#363) Update update-changelog.yml (Blaizzy#364) Update mkdocs configuration (Blaizzy#365) * Update mkdocs config and remove placeholders * Update mkdocs.yml * add overrides Remove hardcoded pytorch install error message (Blaizzy#369) * Remove pytorch install message * Fix formatting * Fix removed change Update processor tests (Blaizzy#370) * update tests for processor * fix tests add jinja to allowed patterns (Blaizzy#376) Fix multi images understand for Qwen2 and Qwen2.5 VL (Blaizzy#377) * fix qwen multi images * fix formatting * remove unsed param * simplify batching logic Qwen2.5 vl fix (Blaizzy#378) * handles case for position ids in qwen2 * remove duplicate * fix: generation command * resets rope_deltas * qwen 2.5vl position embeddings * missing kwarg * rope deltas * fix * nit * address comments * passes image and vdo grid to kwargs * Updated qwen2.5vl * config structure * formatting * fix tests * fix tests and refactor config image ids * fix inference * fix inference * remove unused * fix mask bug * fix offset indexing * Remove unused * remove unused * fix qwen2_5_vl prompt * remove unused --------- Co-authored-by: Prince Verma <princev@lambdatest.com> Co-authored-by: Dillon DuPont <v-ddupont@microsoft.com> Co-authored-by: ddupont <3820588+ddupont808@users.noreply.github.com> Co-authored-by: Prince Verma <prncvrm@gmail.com> Bump mlx to v0.26.0 (Blaizzy#381) bump (Blaizzy#382) fix: Set lora-alpha default to 0.1 for backward consistency fix: Changed to use store_false for consistency with --apply-chat-template parameter fix: Corrected store_false vs store_true
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
add jinja to allowed patterns
Closes #373