update

egs/aishell/paraformer/run.sh

@@ -169,12 +169,8 @@ if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
                 --token_list $token_list \
                 --train_data_path_and_name_and_type ${feats_dir}/${dumpdir}/${train_set}/${scp},speech,${type} \
                 --train_data_path_and_name_and_type ${feats_dir}/${dumpdir}/${train_set}/text,text,text \
-                --train_shape_file ${feats_dir}/asr_stats_fbank_zh_char/${train_set}/speech_shape \
-                --train_shape_file ${feats_dir}/asr_stats_fbank_zh_char/${train_set}/text_shape.char \
                 --valid_data_path_and_name_and_type ${feats_dir}/${dumpdir}/${valid_set}/${scp},speech,${type} \
                 --valid_data_path_and_name_and_type ${feats_dir}/${dumpdir}/${valid_set}/text,text,text \
-                --valid_shape_file ${feats_dir}/asr_stats_fbank_zh_char/${valid_set}/speech_shape \
-                --valid_shape_file ${feats_dir}/asr_stats_fbank_zh_char/${valid_set}/text_shape.char  \
                 --resume true \
                 --output_dir ${exp_dir}/exp/${model_dir} \
                 --config $asr_config \

funasr/utils/prepare_data.py

@@ -162,6 +162,10 @@ def prepare_data(args, distributed_option):
     if args.dataset_type == "large" and args.train_data_file is not None:
         return
     distributed = distributed_option.distributed
+    if not hasattr(args, "train_set"):
+        args.train_set = "train"
+    if not hasattr(args, "dev_set"):
+        args.dev_set = "validation"
     if not distributed or distributed_option.dist_rank == 0:
         filter_wav_text(args.data_dir, args.train_set)
         filter_wav_text(args.data_dir, args.dev_set)