low accuracy on VQAv2 test-std when reproducing prompt tuning experiments

Hi OFA team, lots of appreciation for the great work! Recently i've been trying to reproduce the experiments in the paper OFA-prompt, and i got a base-size model with accuracy of 73.08 on test-dev split. However the test accuracy dropped significantly on test-std set, in which i only got around 22.38(11189 out of 50000). below is my training script:

for total_num_updates in 40000; do
  echo "total_num_updates "${total_num_updates}
  for warmup_updates in 1000; do
    echo "warmup_updates "${warmup_updates}  
    for lr in 0.03; do
      echo "lr "${lr}
      for patch_image_size in 480; do
        echo "patch_image_size "${patch_image_size}

        log_file=${log_dir}/${total_num_updates}"_"${warmup_updates}"_"${lr}"_"${patch_image_size}"_rank"${RANK}".log"
        save_path=${save_dir}/${total_num_updates}"_"${warmup_updates}"_"${lr}"_"${patch_image_size}
        mkdir -p $save_path

        CUDA_VISIBLE_DEVICES=0,1,2,3 python3 -m torch.distributed.launch --nproc_per_node=4 --master_port=${MASTER_PORT} ../../train.py \
            ${data} \
            --selected-cols=${selected_cols} \
            --bpe-dir=${bpe_dir} \
            --user-dir=${user_dir} \
            --restore-file=${restore_file} \
            --reset-optimizer --reset-dataloader --reset-meters \
            --save-dir=${save_path} \
            --task=${task} \
            --arch=${arch} \
            --criterion=${criterion} \
            --label-smoothing=${label_smoothing} \
            --batch-size=${batch_size} \
            --update-freq=${update_freq} \
            --encoder-normalize-before \
            --decoder-normalize-before \
            --share-decoder-input-output-embed \
            --share-all-embeddings \
            --layernorm-embedding \
            --patch-layernorm-embedding \
            --code-layernorm-embedding \
            --resnet-drop-path-rate=${resnet_drop_path_rate} \
            --encoder-drop-path-rate=${encoder_drop_path_rate} \
            --decoder-drop-path-rate=${decoder_drop_path_rate} \
            --dropout=${dropout} \
            --attention-dropout=${attention_dropout} \
            --weight-decay=0.01 \
            --optimizer=adam \
            --adam-betas="(0.9,0.999)" \
            --adam-eps=1e-08 \
            --clip-norm=1.0 \
            --lr-scheduler=polynomial_decay \
            --lr=${lr} \
            --total-num-update=${total_num_updates} \
            --warmup-updates=${warmup_updates} \
            --log-format=simple \
            --log-interval=10 \
            --fixed-validation-seed=7 \
            --keep-last-epochs=15 \
            --save-interval=1 --validate-interval=1 \
            --max-update=${total_num_updates} \
            --best-checkpoint-metric=vqa_score --maximize-best-checkpoint-metric \
            --max-src-length=${max_src_length} \
            --max-object-length=${max_object_length} \
            --max-tgt-length=${max_tgt_length} \
            --find-unused-parameters \
            --freeze-encoder-embedding \
            --freeze-decoder-embedding \
            ${unconstrained_training_flag} \
            --ans2label-file=${ans2label_file} \
            --valid-batch-size=20 \
            --add-type-embedding \
            --scale-attn \
            --scale-fc \
            --encoder-prompt \
            --decoder-prompt \
            --encoder-prompt-type=${prompt_type_method} \
            --decoder-prompt-type=${prompt_type_method} \
            --encoder-prompt-length=${encoder_prompt_length} \
            --decoder-prompt-length=${decoder_prompt_length} \
            --scale-heads \
            --disable-entangle \
            --num-bins=${num_bins} \
            --patch-image-size=${patch_image_size} \
            --prompt-type='none' \
            --fp16 \
            --fp16-scale-window=512 \
            --add-object \
            ${uses_ema} \
            ${store_ema} \
            ${ema_fp32} \
            --ema-decay=${ema_decay} \
            --ema-start-update=${ema_start_update} \
            --val-inference-type=${val_inference_type} \
            --num-workers=0 > ${log_file} 2>&1
      done
    done
  done
done

during evaluating, i changed batch-size to 80 and ran single chunked file in #68 . I wonder whether batch-size and size of test-size affects. Thanks for your attention @JustinLin610

OFA-Sys / OFA

low accuracy on VQAv2 test-std when reproducing prompt tuning experiments #418