Add VANTAGE + TAR reproduction kit for cosmos3 reasoner (vlmevalkit)#245
Open
kding1 wants to merge 1 commit into
Open
Add VANTAGE + TAR reproduction kit for cosmos3 reasoner (vlmevalkit)#245kding1 wants to merge 1 commit into
kding1 wants to merge 1 commit into
Conversation
Signed-off-by: Ke Ding <keding@nvidia.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Adds
evaluation/cosmos3/reasoner/vlmevalkit/— an NVIDIA-customized build of VLMEvalKit that makes the Cosmos3 Reasoner scores on the NVIDIA Metropolis "smart infra" benchmarks publicly reproducible. The reproduction logic lives incosmos_eval/; the rest of the tree is the stock VLMEvalKit inference + scoring engine it drives.Benchmarks shipped (9 total):
AETCBench_all)VANTAGE_VQA,VANTAGE_DVC,VANTAGE_Temporal,VANTAGE_EventVerification,VANTAGE_2DGrounding,VANTAGE_2DPointing,VANTAGE_SOT,VANTAGE_Astro2DWhat's included
cosmos_eval/— the reproduction kit (stdlib-only, no extra eval logic):data/<domain>/<bench>.json— per-benchmark dataset config (model-independent)models/<family>.json— per-family model layer (cosmos)manifest.json— per-benchmarkrun.pyflagsrun_all.py— parallel launcher (compose →run.py→ score)parse_score.py— standalone score reporter that matches the internal pipelinePlus the VLMEvalKit engine (
run.py,vlmeval/, …); the upstream toolkit README is preserved asREADME-vlmevalkit.md.Reproducing scores
Deploy the model behind an OpenAI-compatible endpoint, set the
COSMOS_*env vars, then:python cosmos_eval/run_all.py --model cosmos --concurrency 8 --work-dir ./out Full setup in evaluation/cosmos3/reasoner/vlmevalkit/README.md.