export CUDA_VISIBLE_DEVICES=0
python3 ../generate_passage_embeddings.py \
--model_name_or_path ../../model/contriever-msmarco \
--passages train_robot.jsonl \
--output_dir robot_embeddings \
--shard_id 0 \
--num_shards 1 \
--per_gpu_batch_size 500