wehub-resource-sync
49 lines
1.0 KiB
Bash
49 lines
1.0 KiB
Bash
#!/usr/bin/env bash
|
|
|
|
set -x
|
|
set -e
|
|
|
|
DIR="$( cd "$( dirname "$0" )" && cd .. && pwd )"
|
|
echo "working directory: ${DIR}"
|
|
|
|
MODEL_NAME_OR_PATH=""
|
|
if [[ $# -ge 1 && ! "$1" == "--"* ]]; then
|
|
MODEL_NAME_OR_PATH=$1
|
|
shift
|
|
fi
|
|
|
|
SPLIT="dev"
|
|
if [[ $# -ge 1 && ! "$1" == "--"* ]]; then
|
|
SPLIT=$1
|
|
shift
|
|
fi
|
|
|
|
DEPTH=1000
|
|
# by default, search top-200 for train, top-1000 for dev
|
|
if [ "${SPLIT}" = "train" ]; then
|
|
DEPTH=200
|
|
fi
|
|
|
|
if [ -z "$OUTPUT_DIR" ]; then
|
|
OUTPUT_DIR="${MODEL_NAME_OR_PATH}"
|
|
fi
|
|
if [ -z "$DATA_DIR" ]; then
|
|
DATA_DIR="${DIR}/data/msmarco_bm25_official/"
|
|
fi
|
|
|
|
mkdir -p "${OUTPUT_DIR}"
|
|
|
|
PYTHONPATH=src/ python -u src/inference/search_main.py \
|
|
--model_name_or_path "${MODEL_NAME_OR_PATH}" \
|
|
--search_split "${SPLIT}" \
|
|
--search_batch_size 128 \
|
|
--search_topk "${DEPTH}" \
|
|
--search_out_dir "${OUTPUT_DIR}" \
|
|
--encode_save_dir "${OUTPUT_DIR}" \
|
|
--q_max_len 32 \
|
|
--add_pooler False \
|
|
--dataloader_num_workers 1 \
|
|
--output_dir "/tmp/" \
|
|
--data_dir "${DATA_DIR}" \
|
|
--report_to none "$@"
|