ranker-pointwise-baseline-eval.sh

#!/bin/bash

DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"

model_name="ranker_pointwise_baseline_model"
num_processes=1
gpu_fraction=1.0
device=1
image_model=InceptionV3Model
text_model=LstmModel
match_model=MlpModel
TFRECORD_DIR="${DIR}/data/Ranker_TFRecord_data/d706a6d0dafbc3e2d660cca2f5d5da50b11f181c"

MODEL_DIR="${DIR}/model/${model_name}"
for ckpt in $(ls ${MODEL_DIR} | python ${DIR}/tools/every_n_step.py 20000 | tac); do 
  # the script directory
  VALIDATE_REFERENCE_FILE="${DIR}/model/combination/reference/VALIDATE/reference-13579bdf.json"

  CHECKPOINT_PATH="${MODEL_DIR}/model.ckpt-$ckpt"
  OUTPUT_DIR="${MODEL_DIR}/model.ckpt-${ckpt}.eval"

  mkdir $OUTPUT_DIR

  cd ${DIR}/ranker

  if [ ! -f ${OUTPUT_DIR}/out.csv ]; then
    CUDA_VISIBLE_DEVICES=$device python inference.py \
      --input_file_pattern="${TFRECORD_DIR}/rankertest-VALIDATE-*.tfrecord" \
      --lines_per_image=21 \
      --batch_size=21 \
      --num_readers=1 \
      --checkpoint_path=${CHECKPOINT_PATH} \
      --vocab_file=${DIR}/data/word_counts.txt \
      --output=${OUTPUT_DIR}/out.csv \
      --model=${model} \
      --image_model=${image_model} \
      --text_model=${text_model} \
      --match_model=${match_model} \
      --mlp_num_layers=2 \
      --mlp_num_units="512,256" \
      --mlp_type_activation="sigmoid" \
      --lstm_cell_type="highway" \
      --num_lstm_layers=2 \
      --num_steps=15020
  fi

  if [ ! -f ${OUTPUT_DIR}/out.json ]; then
    cat ${OUTPUT_DIR}/out.csv | python ${DIR}/scripts/ranker_convert_csv_to_oracle_prediction.py > ${OUTPUT_DIR}/out.json
    echo output saved to ${OUTPUT_DIR}/out.json
  fi

  if [ ! -f ${OUTPUT_DIR}/out.eval ]; then
    python ${DIR}/tools/eval/run_evaluations.py --submit ${OUTPUT_DIR}/out.json --ref $VALIDATE_REFERENCE_FILE | tee ${OUTPUT_DIR}/out.eval | grep ^Eval
    echo eval result saved to ${OUTPUT_DIR}/out.eval
  fi

done