feat: synthetic image and video data generation for VLM benchmarking#732

Open

zakariaelh wants to merge 14 commits into

vllm-project:mainfrom

zakariaelh:feat/synthetic-multimodal

Commits on Jun 25, 2026

extras/vision: add synthesize_image and synthesize_video helpers

authored and
alityb
committed
data: add synthetic_image and synthetic_video deserializers

authored and
alityb
committed
tests: unit + integration coverage for synthetic_image and synthetic_video

authored and
alityb
committed
docs: README usage examples for synthetic_image and synthetic_video

authored and
alityb
committed
fix: declare image/video features and make encoders idempotent

zakariaelh
authored and
alityb
committed
tests: add WRITTEN BY AI marker per AGENTS.md

zakariaelh
authored and
alityb
committed
Fix pre-existing lint and type-check failures

Jckwind
authored and
alityb
committed
Add coordinate warp to synthetic gradient generator

Jckwind
authored and
alityb
committed
docs: move synthetic visual data out of README into dedicated guide

zakariaelh
authored and
alityb
committed
review: address second-pass docs and code comments

zakariaelh
authored and
alityb
committed
data: align synthetic vision with kind routing
alityb
committed
tests: update synthetic vision kind coverage
alityb
committed
docs: update synthetic vision CLI examples
alityb
committed
tox: install dev extra for quality envs
alityb
committed