take away performance enhancement - #169
Merged
Merged
Conversation
Align take-away with dine-in so both applications share a single OVMS model instance instead of loading Qwen2.5-VL-7B and MiniCPM separately. Configuration: - OVMS_MODEL_NAME defaults to openbmb/MiniCPM-V-4_5-int4, VLM_PRECISION to int4, across .env.example, Makefile, docker-compose, application.yaml and the Python fallbacks in src/. - docker-compose now references the locally built images by :latest via overridable *_IMAGE variables. setup_models.sh: - Per-app default model; --app take-away and --app dine-in both resolve to MiniCPM-V-4_5-int4 when OVMS_MODEL_NAME is unset. - VLM_PRECISION is derived from the precision suffix of OVMS_MODEL_NAME and warns when the two disagree, since the suffix is only a directory label and an int8 export would otherwise land in an "-int4" directory. - Legacy Qwen model migration is restricted to the Qwen target path so Qwen weights can no longer be copied into a directory named for another model. VLM client: - Send chat_template_kwargs.enable_thinking=false. MiniCPM-V-4.5 is a hybrid reasoning model whose chat template otherwise opens a <think> block that consumes the whole max_completion_tokens budget, truncating the item list. Any stray <think> output is stripped before parsing. - Fit frames into a 448x448 square canvas preserving aspect ratio (LANCZOS, white padding, light contrast/sharpening), matching the dine-in pipeline and replacing a fixed 512x512 resize that distorted proportions. Measured on the sample video (1 worker, GPU): average VLM latency dropped from 8657 ms to ~1900 ms and completion tokens from a truncated 100 to 16-18 clean "item x qty" lines. Also removes "water bottle" from order 651 in config/orders.json, as it does not appear in the sample video and was reported as a false mismatch. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
sainijit
requested review from
sachinkaushik and
sumanaga
and removed request for
sachinkaushik
July 30, 2026 15:07
avinash-palleti
approved these changes
Jul 30, 2026
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
3 participants
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.
No description provided.