AssertionError in eval.py

[RinuJoseph]

When running the eval.py script for evaluation using the provided command, I encounter an AssertionError. It seems related to the transform parameter in the dataset initialization. This happens only when I use sam2 model_type. It works fine for "ori" model_type

torchrun --standalone --nproc_per_node=2 eval.py
    --version YxZhang/evf-sam2
    --dataset_dir ./dataset
    --val_dataset "refcoco|unc|val"
    --model_type sam2

[CoderZhangYx]

edit eval.py:
addfrom utils.dataset import Resize
and add transfrom=ResizeLongestSide(1024) if args.model_type=="ori" else Resize(1024) at

EVF-SAM/eval.py

Lines 110 to 111 in 0b3bc4c

	model_type=args.model_type
	)

Sorry for our carelessness‌.

[RinuJoseph]

Thank you for the response. One question: what version should be used for model_type "effi"?

[CoderZhangYx]

It's EVF-SAM based on Efficient-SAM. But we haven't released the checkpoint.

[RinuJoseph]

I see. Thank you for clarifying