search

alumae / kaldi-gstreamer-server

Real-time full-duplex speech recognition server, based on the Kaldi toolkit and the GStreamer framwork.
BSD 2-Clause "Simplified" License
1.07k stars 341 forks source link

decoder with CSJ -> worker: segmentation fault (core dumped) #247

Open averak opened 3 years ago

averak commented 3 years ago

Hi!

I'm beginner of Kaldi and Gstreamserver. I have a trained kaldi model with nnet3 on CSJ. That's why I tried to run this model with docker-kaldi-gstreamer-server.

  1. start master & worker server 
    $ python kaldigstserver/master_server.py --port=80
$ python kaldigstserver/worker.py -u ws://localhost:80/worker/ws/speech -c csj_nnet3.yaml

(The details of the yaml file are written at the end of this issue)

  1. run client.py 
    $ python kaldigstserver/client.py -u ws://localhost:80/client/ws/speech -r 32000 test/data/english_test.raw

Then, the worker server went down leaving the following log.

INFO 2020-12-30 13:59:01,256 101 GET /client/ws/speech?content-type= (127.0.0.1) 0.41ms
INFO 2020-12-30 13:59:01,257 dd1a08a3-930b-489d-a176-10a21f2325fb: OPEN
INFO 2020-12-30 13:59:01,257 dd1a08a3-930b-489d-a176-10a21f2325fb: Request arguments: content-type=""
INFO 2020-12-30 13:59:01,258 dd1a08a3-930b-489d-a176-10a21f2325fb: Using worker <__main__.DecoderSocketHandler object at 0x7f681fb2e150>
2020-12-30 13:59:01 -   DEBUG:   __main__: <undefined>: Got message from server of type <class 'ws4py.messaging.TextMessage'>
2020-12-30 13:59:01 -    INFO:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Initializing request
2020-12-30 13:59:01 -    INFO:   __main__: dd1a08a3-930b-489d-a176-10a21f2325fb: Started timeout guard
2020-12-30 13:59:01 -    INFO:   __main__: dd1a08a3-930b-489d-a176-10a21f2325fb: Initialized request
2020-12-30 13:59:01 -   DEBUG:   __main__: dd1a08a3-930b-489d-a176-10a21f2325fb: Checking that decoder hasn't been silent for more than 10 seconds
INFO 2020-12-30 13:59:01,483 dd1a08a3-930b-489d-a176-10a21f2325fb: Forwarding client message (<type 'str'>) of length 2048 to worker
2020-12-30 13:59:01 -   DEBUG:   __main__: dd1a08a3-930b-489d-a176-10a21f2325fb: Got message from server of type <class 'ws4py.messaging.BinaryMessage'>
2020-12-30 13:59:01 -   DEBUG:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Pushing buffer of size 2048 to pipeline
2020-12-30 13:59:01 -   DEBUG:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Pushing buffer done
INFO 2020-12-30 13:59:01,741 dd1a08a3-930b-489d-a176-10a21f2325fb: Forwarding client message (<type 'str'>) of length 2048 to worker
2020-12-30 13:59:01 -   DEBUG:   __main__: dd1a08a3-930b-489d-a176-10a21f2325fb: Got message from server of type <class 'ws4py.messaging.BinaryMessage'>
2020-12-30 13:59:01 -   DEBUG:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Pushing buffer of size 2048 to pipeline
2020-12-30 13:59:01 -   DEBUG:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Pushing buffer done
INFO 2020-12-30 13:59:01,995 dd1a08a3-930b-489d-a176-10a21f2325fb: Forwarding client message (<type 'str'>) of length 2048 to worker
2020-12-30 13:59:01 -   DEBUG:   __main__: dd1a08a3-930b-489d-a176-10a21f2325fb: Got message from server of type <class 'ws4py.messaging.BinaryMessage'>
2020-12-30 13:59:01 -   DEBUG:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Pushing buffer of size 2048 to pipeline
2020-12-30 13:59:01 -   DEBUG:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Pushing buffer done
INFO 2020-12-30 13:59:02,250 dd1a08a3-930b-489d-a176-10a21f2325fb: Forwarding client message (<type 'str'>) of length 2048 to worker
2020-12-30 13:59:02 -   DEBUG:   __main__: dd1a08a3-930b-489d-a176-10a21f2325fb: Got message from server of type <class 'ws4py.messaging.BinaryMessage'>
2020-12-30 13:59:02 -   DEBUG:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Pushing buffer of size 2048 to pipeline
2020-12-30 13:59:02 -   DEBUG:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Pushing buffer done
2020-12-30 13:59:02 -   DEBUG:   __main__: dd1a08a3-930b-489d-a176-10a21f2325fb: Checking that decoder hasn't been silent for more than 10 seconds
INFO 2020-12-30 13:59:02,504 dd1a08a3-930b-489d-a176-10a21f2325fb: Forwarding client message (<type 'str'>) of length 2048 to worker
2020-12-30 13:59:02 -   DEBUG:   __main__: dd1a08a3-930b-489d-a176-10a21f2325fb: Got message from server of type <class 'ws4py.messaging.BinaryMessage'>
2020-12-30 13:59:02 -   DEBUG:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Pushing buffer of size 2048 to pipeline
2020-12-30 13:59:02 -   DEBUG:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Pushing buffer done
2020-12-30 13:59:02 -    INFO:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Connecting audio decoder
2020-12-30 13:59:02 -    INFO:   decoder2: dd1a08a3-930b-489d-a176-10a21f2325fb: Connected audio decoder
INFO 2020-12-30 13:59:02,645 Worker <__main__.WorkerSocketHandler object at 0x7f681fb86290> leaving
INFO 2020-12-30 13:59:02,649 dd1a08a3-930b-489d-a176-10a21f2325fb: Handling on_connection_close()
INFO 2020-12-30 13:59:02,649 dd1a08a3-930b-489d-a176-10a21f2325fb: Closing worker connection
[2]    8887 segmentation fault (core dumped)  python kaldigstserver/worker.py -u ws://localhost:80/worker/ws/speech -c

My yaml file is below.

use-nnet2: True
decoder:
    nnet-mode: 3
    use-threaded-decoder:  true
    model : /opt/models/tdnn1a_online/final.mdl
    word-syms : /opt/models/tdnn1a_online/phones.txt
    mfcc-config : /opt/models/tdnn1a_online/conf/mfcc.conf
    ivector-extraction-config : /opt/models/tdnn1a_online/conf/ivector_extractor.conf
    max-active: 10000
    beam: 10.0
    lattice-beam: 6.0
    acoustic-scale: 0.083
    do-endpointing : true
    endpoint-silence-phones : "1:2:3:4:5:6:7:8:9:10"
    traceback-period-in-secs: 0.25
    chunk-length-in-secs: 0.25
    num-nbest: 1
out-dir: tmp

use-vad: False
silence-timeout: 10

# Just a sample post-processor that appends "." to the hypothesis
post-processor: perl -npe 'BEGIN {use IO::Handle; STDOUT->autoflush(1);} sleep(1); s/(.*)/\1./;'

#post-processor: (while read LINE; do echo $LINE; done)

# A sample full post processor that add a confidence score to 1-best hyp and deletes other n-best hyps
#full-post-processor: ./sample_full_post_processor.py

logging:
    version : 1
    disable_existing_loggers: False
    formatters:
        simpleFormater:
            format: '%(asctime)s - %(levelname)7s: %(name)10s: %(message)s'
            datefmt: '%Y-%m-%d %H:%M:%S'
    handlers:
        console:
            class: logging.StreamHandler
            formatter: simpleFormater
            level: DEBUG
    root:
        level: DEBUG
        handlers: [console]

Thanks

sirifarif commented 3 years ago

Did you ever use it successfully for some recognition? or It's just at the beginning of your setup.

hezhenke commented 2 years ago

i met the same problem