| | |
| | | stage=1 |
| | | stop_stage=3 |
| | | |
| | | bert_model_root="../../huggingface_models" |
| | | bert_model_name="bert-base-chinese" |
| | | raw_dataset_path="../DATA" |
| | | model_path=${bert_model_root}/${bert_model_name} |
| | | nj=64 |
| | | model_path=${bert_model_name} |
| | | |
| | | . utils/parse_options.sh || exit 1; |
| | | |
| | | nj=32 |
| | | |
| | | for data_set in train dev test;do |
| | | scp=$raw_dataset_path/data/${data_set}/text |
| | |
| | | cat ${local_records_dir}/embeds.${JOB}.shape || exit 1; |
| | | done > ${local_scp_dir_raw}/embeds.shape |
| | | fi |
| | | |
| | | cp ${local_scp_dir_raw}/embeds.scp ${raw_dataset_path}/data/${data_set}/embeds.scp |
| | | done |
| | | |
| | | echo "embeds is in: ${local_scp_dir_raw}" |