update repo

egs/aishell/data2vec_paraformer_finetune/conf/train_asr_paraformer_transformer_12e_6d_3072_768.yaml

@@ -105,6 +105,8 @@ predictor_conf:
   r_order: 1
 
 dataset_conf:
+    data_names: speech,text
+    data_types: sound,text
     shuffle: True
     shuffle_conf:
         shuffle_size: 2048

egs/aishell/data2vec_transformer_finetune/conf/train_asr_transformer_12e_6d_3072_768.yaml

@@ -96,6 +96,8 @@ specaug_conf:
     num_time_mask: 2
 
 dataset_conf:
+    data_names: speech,text
+    data_types: sound,text
     shuffle: True
     shuffle_conf:
         shuffle_size: 2048

egs/aishell/paraformer/conf/train_asr_paraformer_conformer_12e_6d_2048_256.yaml

@@ -93,6 +93,8 @@ predictor_conf:
     tail_threshold: 0.45
 
 dataset_conf:
+    data_names: speech,text
+    data_types: sound,text
     shuffle: True
     shuffle_conf:
         shuffle_size: 2048

egs/aishell/paraformerbert/local/extract_embeds.sh

@@ -54,6 +54,8 @@ for data_set in train dev test;do
             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}"

egs/aishell/transformer/conf/train_asr_transformer.yaml

@@ -73,6 +73,8 @@ scheduler_conf:
     warmup_steps: 25000
 
 dataset_conf:
+    data_names: speech,text
+    data_types: sound,text
     shuffle: True
     shuffle_conf:
         shuffle_size: 2048

egs/aishell2/conformer/conf/train_asr_conformer.yaml

@@ -84,6 +84,7 @@ specaug_conf:
     num_time_mask: 2
 
 dataset_conf:
+    data_names: speech,text
     data_types: sound,text
     shuffle: True
     shuffle_conf:

egs/aishell2/paraformer/conf/train_asr_paraformer_conformer_20e_1280_320_6d_1280_320.yaml

@@ -94,6 +94,7 @@ predictor_conf:
   r_order: 1
 
 dataset_conf:
+    data_names: speech,text
     data_types: sound,text
     shuffle: True
     shuffle_conf:

egs/aishell2/paraformerbert/local/extract_embeds.sh

@@ -54,6 +54,8 @@ for data_set in train dev_ios;do
             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}"

egs/aishell2/transformer/conf/train_asr_transformer.yaml

@@ -78,6 +78,7 @@ specaug_conf:
     num_time_mask: 2
 
 dataset_conf:
+    data_names: speech,text
     data_types: sound,text
     shuffle: True
     shuffle_conf: