apt-get update
apt-get install openjdk-8-jdk
pip install evaluate sacrebleu==1.5.1
pip install pycocoevalcap
pip install edit_distance editdistance
mkdir -p eval_audio/caption_evaluation_tools
git clone https://github.com/audio-captioning/caption-evaluation-tools.git eval_audio/caption_evaluation_tools
cd eval_audio/caption_evaluation_toolscoco_caption/
./get_stanford_models.sh
cd ../..
pip install sacrebleu
pip install sacrebleu\[ja\]
pip install sed_eval
pip install dcase_util

mkdir -p data/asr && cd data/asr

# download audios from above links

# download converted files
wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/evaluation/librispeech_eval.jsonl
wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/evaluation/aishell1_eval.jsonl
wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/evaluation/aishell2_eval.jsonl


cd ../..

 for ds in "librispeech" "aishell1" "aishell2"
 do
     python -m torch.distributed.launch --use_env \
         --nproc_per_node ${NPROC_PER_NODE:-8} --nnodes 1 \
         evaluate_asr.py \
         --checkpoint $checkpoint \
         --dataset $ds \
         --batch-size 20 \
         --num-workers 2
 done

mkdir -p data/st && cd data/st

# download audios from https://commonvoice.mozilla.org/en/datasets

# download converted files
wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/evaluation/covost2_eval.jsonl

cd ../..

ds="covost2"
python -m torch.distributed.launch --use-env \
    --nproc_per_node ${NPROC_PER_NODE:-8} --nnodes 1 \
    evaluate_st.py \
    --checkpoint $checkpoint \
    --dataset $ds \
    --batch-size 8 \
    --num-workers 2 

{"audio": "001001.wav", 
  "gt": "<|0.00|><|sil|><|0.54|>面<|0.86|>条<|1.17|>机<|1.46|><|sil|><|2.01|>", 
  "source": "aishell2"}

ds="covost2"
python -m torch.distributed.launch --use-env \
    --nproc_per_node ${NPROC_PER_NODE:-8} --nnodes 1 \
    evaluate_st.py \
    --checkpoint $checkpoint \
    --dataset $ds \
    --batch-size 8 \
    --num-workers 2 

mkdir -p data/caption && cd data/caption

# download audios from https://zenodo.org/record/3490684

# download converted files
wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/evaluation/clotho_eval.jsonl

cd ../..

ds="clotho"
python -m torch.distributed.launch --use-env \
    --nproc_per_node ${NPROC_PER_NODE:-8} --nnodes 1 \
    evaluate_caption.py \
    --checkpoint $checkpoint \
    --dataset $ds \
    --batch-size 8 \
    --num-workers 2 \

mkdir -p data/asc && cd data/asc

# download TUT2017 datasets from https://zenodo.org/records/1040168
# download CochlScene datasets from https://zenodo.org/records/7080122


# download converted files
wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/evaluation/tut2017_eval.jsonl
wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/evaluation/cochlscene_eval.jsonl

cd ../..

 for ds in "cochlscene" "tut2017"
 do
     python -m torch.distributed.launch --use_env \
         --nproc_per_node ${NPROC_PER_NODE:-8} --nnodes 1 \
         evaluate_scene.py \
         --checkpoint $checkpoint \
         --dataset $ds \
         --batch-size 20 \
         --num-workers 2
 done

mkdir -p data/ser && cd data/ser

# download MELD datasets from above link

# download converted files
wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/evaluation/meld_eval.jsonl


cd ../..

ds="meld"
python -m torch.distributed.launch --use-env \
    --nproc_per_node ${NPROC_PER_NODE:-8} --nnodes 1 \
    evaluate_emotion.py \
    --checkpoint $checkpoint \
    --dataset $ds \
    --batch-size 8 \
    --num-workers 2 \
    --few-shot 0

mkdir -p data/aqa && cd data/aqa

# download Clotho-AQA dataset from https://zenodo.org/records/6473207
# download converted files
wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/evaluation/clothoaqa_eval.jsonl


cd ../..

ds="clothoaqa"
python -m torch.distributed.launch --use-env \
    --nproc_per_node ${NPROC_PER_NODE:-8} --nnodes 1 \
    evaluate_aqa.py \
    --checkpoint $checkpoint \
    --dataset $ds \
    --batch-size 8 \
    --num-workers 2 

mkdir -p data/vsc && cd data/vsc

# download dataset from the above link
# download converted files
wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/evaluation/vocalsound_eval.jsonl


cd ../..

ds="clothoaqa"
python -m torch.distributed.launch --use-env \
    --nproc_per_node ${NPROC_PER_NODE:-8} --nnodes 1 \
    evaluate_aqa.py \
    --checkpoint $checkpoint \
    --dataset $ds \
    --batch-size 8 \
    --num-workers 2 

mkdir -p data/mna && cd data/mna

# download NSynth datasets from above link

# download converted files
wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/evaluation/nsynth_eval.jsonl

cd ../..

ds="nsynth"
python -m torch.distributed.launch --use-env \
    --nproc_per_node ${NPROC_PER_NODE:-8} --nnodes 1 \
    evaluate_note_analysis.py \
    --checkpoint $checkpoint \
    --dataset $ds \
    --batch-size 8 \
    --num-workers 2