fix typo.

examples/hi_xiaowen/s0/conf/ds_tcn_ctc_base.yaml

@@ -44,7 +44,7 @@ optim_conf:
 
 training_config:
     grad_clip: 5
-    max_epoch: 30
+    max_epoch: 50
     log_interval: 100
     criterion: ctc
 

examples/hi_xiaowen/s0/run_ctc.sh

@@ -11,15 +11,19 @@ num_keywords=2599
 config=conf/ds_tcn_ctc.yaml
 norm_mean=true
 norm_var=true
-gpus="4,5,6,7"
+gpus="0,1,2,3"
 
 checkpoint=
 dir=exp/ds_tcn_ctc_ft
-
+average_model=true
 num_average=30
-score_checkpoint=$dir/avg_${num_average}.pt
+if $average_model ;then
+  score_checkpoint=$dir/avg_${num_average}.pt
+else
+  score_checkpoint=$dir/final.pt
+fi
 
-download_dir=/data/megastore/Datasets/ASR/KWS/nihaoxiaowen # your data dir
+download_dir=/mnt/52_disk/back/DuJing/data/nihaowenwen # your data dir
 
 . tools/parse_options.sh || exit 1;
 window_shift=50
@@ -100,7 +104,7 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ] && [ $trainbase == true ]; then
       exit
     fi
     if [ ! -f $trainbase_dir/$x/wav.dur ]; then
-      tools/wav_to_duration.sh --nj 8 $trainbase_dir/$x/wav.scp $trainbase_dir/$x/wav.dur
+      tools/wav_to_duration.sh --nj 128 $trainbase_dir/$x/wav.scp $trainbase_dir/$x/wav.dur
     fi
 
     # Here we use tokens.txt and lexicon.txt to convert txt into index
@@ -124,12 +128,13 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ] && [ $trainbase == true ]; then
       --train_data $trainbase_dir/train/data.list \
       --cv_data $trainbase_dir/dev/data.list \
       --model_dir $trainbase_exp \
-      --num_workers 8 \
+      --num_workers 2 \
+      --ddp.dist_backend nccl \
       --num_keywords $num_keywords \
       --min_duration 50 \
       --seed 666 \
-      $cmvn_opts \
+      $cmvn_opts # \
-      --checkpoint $trainbase_exp/23.pt
+      #--checkpoint $trainbase_exp/23.pt
 fi
 
 if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
@@ -167,16 +172,19 @@ fi
 
 if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
   echo "Do model average, Compute FRR/FAR ..."
-  python wekws/bin/average_model.py \
+  if $average_model; then
-    --dst_model $score_checkpoint \
+    python wekws/bin/average_model.py \
-    --src_path $dir  \
+      --dst_model $score_checkpoint \
-    --num ${num_average} \
+      --src_path $dir  \
-    --val_best
+      --num ${num_average} \
+      --val_best
+  fi
   result_dir=$dir/test_$(basename $score_checkpoint)
   mkdir -p $result_dir
   python wekws/bin/score_ctc.py \
     --config $dir/config.yaml \
     --test_data data/test/data.list \
+    --gpu 0  \
     --batch_size 256 \
     --checkpoint $score_checkpoint \
     --score_file $result_dir/score.txt  \

wekws/bin/compute_det_ctc.py

@@ -17,7 +17,7 @@
 import argparse, logging, glob
 import json, re, os, numpy as np
 import matplotlib.pyplot as plt
-import pypinyin
+import pypinyin  # for Chinese Character
 
 def split_mixed_label(input_str):
     tokens = []
@@ -165,13 +165,13 @@ if __name__ == '__main__':
     parser.add_argument(
         '--xlim',
         type=int,
-        default=5,
+        default=10,
         help='xlim：range of x-axis, x is false alarm per hour')
     parser.add_argument('--x_step', type=int, default=1, help='step on x-axis')
     parser.add_argument(
         '--ylim',
         type=int,
-        default=75,
+        default=100,
         help='ylim：range of y-axis, y is false rejection rate')
     parser.add_argument('--y_step', type=int, default=5, help='step on y-axis')
 
@@ -228,7 +228,7 @@ if __name__ == '__main__':
                 false_alarm_rate = num_false_alarm / filler_num
 
                 fout.write('{:.3f} {:.6f} {:.6f}\n'.format(
-                    threshold, false_alarm_per_hour, threshold))
+                    threshold, false_alarm_per_hour, false_reject_rate))
                 threshold += args.step
     if args.det_curve_path :
         det_curve_path = args.det_curve_path

wekws/bin/score_ctc.py

@@ -192,10 +192,10 @@ def main():
                 if hit_keyword is not None:
                     # fout.write('{} detected [{:.2f} {:.2f}] {} {:.3f}\n'\
                     #          .format(key, start*0.03, end*0.03, hit_keyword, hit_score))
-                    fout.write('{} {} {:.3f}\n'.format(
+                    fout.write('{} detected {} {:.3f}\n'.format(
                         key, hit_keyword, hit_score))
                 else:
-                    fout.write('{} -1 -1\n'.format(key))
+                    fout.write('{} rejected\n'.format(key))
 
             if batch_idx % 10 == 0:
                 print('Progress batch {}'.format(batch_idx))