Retriever/bash_retr.sh

#!/bin/bash

export CUDA_VISIBLE_DEVICES=0


QUESTION_FILE="../data/webq-test.txt" # 问题测试集
EMBEDDING_FILE="../corpus/ctx_embeddings.pkl" # 语料库的embedding
INDEX_PATH="../index/webq_index" # 语料库的faiss索引
OUTPUT_FILE="../output/result.pkl" # 检索结果输出
BATCH_SIZE=32
MODEL_NAME="facebook/dpr-question_encoder-multiset-base" # 将问题编码为embedding的模型
DEVICE="cuda"

python retriever.py \
    --questions_file $QUESTION_FILE \
    --context_embeddings_file $EMBEDDING_FILE \
    --model_name $MODEL_NAME \
    --index_path $INDEX_PATH\
    --output_file $OUTPUT_FILE \
    --batch_size $BATCH_SIZE \
    --top_docs 100 \
    --device $DEVICE