UCF performance

[xiaoxinning]

I'm very sorry to bother you.actually I don't know if I missed any crucial steps.I just using you extracted temporal supervised (C3D/VGGish) features and run the scripts but my result gets really low, may I ask you if it's gets right ,every time I run should be close to your score? also, if I using your trained weights,I can get the same performance as you. actually I supposed it can't be the problems of the packages because I'm using the same package version to run AVCA and gets correct.
this is the scripts I use : python3 main.py --cfg /root/autodl-tmp/TCAF-GZSL/config/best/best_cls/best_ucf.yaml --root_dir avgzsl_benchmark_non_averaged_datasets/UCF --log_dir networks/ucf_cls --dataset_name UCF --run all
performance GZSL performance=27.99, ZSL performance=28.42