Hugging Face's logo

Spaces:
LanguageBind
/
Video-LLaVA
Running on A10G

App Files Community
4
Video-LLaVA
File size: 1,391 Bytes 
5244231
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
 
#!/bin/bash

gpu_list="${CUDA_VISIBLE_DEVICES:-0}"
IFS=',' read -ra GPULIST <<< "$gpu_list"

CHUNKS=${#GPULIST[@]}

CKPT_NAME="Video-LLaVA-7B"
CKPT="checkpoints/${CKPT_NAME}"
SPLIT="llava_gqa_testdev_balanced"
EVAL="eval"
GQADIR="${EVAL}/gqa/data"

for IDX in $(seq 0 $((CHUNKS-1))); do
    CUDA_VISIBLE_DEVICES=${GPULIST[$IDX]} python3 -m llava.eval.model_vqa_loader \
        --model-path ${CKPT} \
        --question-file ${EVAL}/gqa/$SPLIT.jsonl \
        --image-folder ${EVAL}/gqa/data/images \
        --answers-file ${EVAL}/gqa/answers/$SPLIT/${CKPT_NAME}/${CHUNKS}_${IDX}.jsonl \
        --num-chunks $CHUNKS \
        --chunk-idx $IDX \
        --temperature 0 \
        --conv-mode vicuna_v1 &
done

wait

output_file=${EVAL}/gqa/answers/$SPLIT/${CKPT_NAME}/merge.jsonl

# Clear out the output file if it exists.
> "$output_file"

# Loop through the indices and concatenate each file.
for IDX in $(seq 0 $((CHUNKS-1))); do
    cat ${EVAL}/gqa/answers/$SPLIT/${CKPT_NAME}/${CHUNKS}_${IDX}.jsonl >> "$output_file"
done

mkdir -p $GQADIR/$SPLIT/${CKPT_NAME}
python3 scripts/convert_gqa_for_eval.py --src $output_file --dst $GQADIR/$SPLIT/${CKPT_NAME}/testdev_balanced_predictions.json

cd $GQADIR
python3 eval/eval_gqa.py --tier $SPLIT/${CKPT_NAME}/testdev_balanced \
                         --questions /scc_cephfs/yy/lb/LLaVA-Video-YY/questions1.2/testdev_balanced_questions.json