Unable to use audio features properly

[Psycho-9]

Hello, I downloaded the QVHighlights dataset from https://github.com/TencentARC/UMT and found that there is only one more folder called pann_feature than the original features. Can this be considered as an additional audio feature? I added the folder pann_feature as an audio feature in the original train script, but the effect is not good. How can I use the audio feature reasonably?
Below is my train_audio script：

if [[ ${v_feat_types} == *"slowfast"* ]]; then
  v_feat_dirs+=(${feat_root}/slowfast_features)
  (( v_feat_dim += 2304 ))  # double brackets for arithmetic op, no need to use ${v_feat_dim}
fi
if [[ ${v_feat_types} == *"clip"* ]]; then
  v_feat_dirs+=(${feat_root}/clip_features)
  (( v_feat_dim += 512 ))
fi

if [[ ${t_feat_type} == "clip" ]]; then
  t_feat_dir=${feat_root}/clip_text_features/
  t_feat_dim=512
else
  echo "Wrong arg for t_feat_type."
  exit 1
fi

if [[ ${a_feat_type} == "pann" ]]; then
  a_feat_dir=${feat_root}/pann_features/
  a_feat_dim=2050
else
  echo "Wrong arg for t_feat_type."
  exit 1
fi