extract perception_test audio feature problems

[WannaSir]

i use following command to extract audio features:

export CUDA_VISIBLE_DEVICES=6,7
python tools/run_net.py --cfg /data/nvme2/dky/lc/TIM/feature_extractors/auditory_slowfast/configs/PERCEPTION/SLOWFAST_R50.yaml \
NUM_GPUS 2 \
OUTPUT_DIR /data/hdd/lishenshen/timfeature_data/perception_test-feature \
PERCEPTION.AUDIO_DATA_DIR /data/hdd/lishenshen/data/perception_test/valid_audios/audios \
PERCEPTION.TEST_LIST /data/nvme2/dky/lc/TIM/annotations/Perception_Test/Perception_Test_1_second_validation_feature_times.pkl \
TEST.CHECKPOINT_FILE_PATH /data/nvme2/dky/lc/TIM/feature_extractors/auditory_slowfast/pretrained_models/asf_vggsound.pyth \
TEST.NUM_FEATURES 1

and then obtain files: feature.npy and metadata.npy, then i post-process the features with numpy files using the follow command: python utils/make_npyfiles.py --feature_file /data/hdd/lishenshen/timfeature_data/perception_test-feature/features.npy --pickle_file /data/nvme2/dky/lc/TIM/annotations/Perception_Test/Perception_Test_1_second_validation_feature_times.pkl --out_dir /data/hdd/lishenshen/timfeature_data/perception_test-feature/val_a

but there will be an error : File "/data/nvme2/dky/lc/TIM/feature_extractors/auditory_slowfast/utils/make_npyfiles.py", line 26, in main assert len(feature) == len(feature_pandas) AssertionError but i do not konw how to fix it ,thank you!

[WannaSir]

and the length of them are above

[JacobChalk]

The length of the feature array should indeed be 593655, the same as the pandas data frame.

The extracted features were of size 57 due to an old piece of residual test code left over. We have fixed this now so that the data loader will construct properly and extract all of the videos, rather than a single specific video. Apologies for the inconvenience!