MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

As synced, lets convert this PR to draft and once ready with test actions, we can merge this PR.

@anandhu-eng good to merge?

Hi @arjunsuresh , The Bevformer CPU workflow is running fine. I m getting an accuracy of 0.268 mAP which is comparable to the result taken by Rod - 0.2683556 mAP. There is some issue in summary table - results not getting populated in latency and accuracy columns. We can sync and raise that as a separate PR.

Edit: The following run cmd was used in testing:

mlc run script --tags=run-abtf-inference,reference,_v0.5,_full --model=bevformer --docker --docker_cache=no --quiet --env.MLC_USE_DATASET_FROM_HOST=yes --env.MLC_USE_MODEL_FROM_HOST=yes --device=cpu --implementation=reference --framework=onnxruntime --scenario=SingleStream --batch_size=1