search

modelscope / FunASR

A Fundamental End-to-End Speech Recognition Toolkit and Open Source SOTA Pretrained Models, Supporting Speech Recognition, Voice Activity Detection, Text Post-processing etc.
https://www.funasr.com
Other
6.19k stars 659 forks source link

Issue with Stage 1: Feature and CMVN Generation in Speech Recognition script #1058

Open qiulin13145 opened 11 months ago

qiulin13145 commented 11 months ago

Issue: I'm following the instructions from the "Speech Recognition" documentation and I'm facing an issue when executing bash run.sh. I was able to pass stage 0 successfully, but the script seems to get stuck at stage 1.

Steps to reproduce:

  1. Run bash run.sh

Expected result: The script should proceed past stage 1.

Observed result: The script gets stuck at stage 1 with no further output. Upon investigation, I found that the script is stuck at line 311 in run.pl, and the corresponding subprocess remains in the 'S' (Sleeping) state. 1

image

image

Question: I'm unsure about the cause of this issue. I would appreciate any guidance or help in resolving it. Thanks in advance!

LauraGPT commented 11 months ago

Maybe you could change the code and try it again, in egs/aishell/paraformer/utils/apply_cmvn.sh line 20:

#$cmd JOB=1:$nj $logdir/apply_cmvn.JOB.log \
#    python utils/apply_cmvn.py -a $fbankdir/ark/feats.JOB.ark \
#        -c $cmvn_file -i JOB -o ${dump_dir} \
#        || exit 1;

for JOB in `seq 1 $nj`;do
  {
      python utils/apply_cmvn.py -a $fbankdir/ark/feats.${JOB}.ark \
    -c $cmvn_file -i ${JOB} -o ${dump_dir} || exit 1;
  } &> $logdir/apply_cmvn.${JOB}.log &
done
qiulin13145 commented 11 months ago

Thank you for your response. Unfortunately, despite my attempts, I'm still running into the same issues and errors as before. Do you have any other suggestions or solutions?