how to generate pkl file

[bigdata2013lee]

hi, thx for ur share this project, in remembr/scripts/preprocess_captions.py: # load folders
pkl_files = glob.glob(os.path.join(args.data_path, str(SEQUENCE_ID), '*.pkl')).

i wanna to know, how to create this pkl files ???

[AbrarAnwar]

Follow the instructions in eval.md. We recently had them updated.

This requires the CODa dataset to be downloaded first. Here is an example script:

python scripts/preprocess_captions.py \
    --seq_id 0 \
    --seconds_per_caption 3 \
    --model-path Efficient-Large-Model/VILA1.5-13b
    --captioner_name VILA1.5-13b 
    --out_path data/captions/0/captions

These pkl files are created in the preprocess_coda_all.sh file, also described in eval.md.