running ASpIRE Chain Model

[prokopevaleksey]

trafficstars

Hi, first of all thank you a lot for this project!

It would be just awesome to configure this docker for "state-of-the-art" ASpIRE Chain Model. I'm trying to configure it but so far it's quite confusing for me.

Any tips for that?

The best wishes, Aleksei

[jcsilva]

Hi,

if I'm not wrong, Chain Models use NNET3 framework. So, you could use something like this:

use-nnet2: True decoder: # All the properties nested here correspond to the kaldinnet2onlinedecoder GStreamer plugin properties. # Use gst-inspect-1.0 ./libgstkaldionline2.so kaldinnet2onlinedecoder to discover the available properties nnet-mode: 3 use-threaded-decoder: True model : /path/to/final.mdl word-syms : /path/to/words.txt fst : /path/to/HCLG.fst mfcc-config : /path/to/mfcc.conf ivector-extraction-config : /path/to/ivector_extractor.conf max-active: 10000 beam: 15.0 lattice-beam: 8.0 acoustic-scale: 0.06 do-endpointing : false ##endpoint-silence-phones : "1:2:3:4:5:6:7:8:9:10" traceback-period-in-secs: 1 chunk-length-in-secs: 1 num-nbest: 10 #Additional functionality that you can play with: #lm-fst: /opt/models/nnet_ms_online/G.fst #big-lm-const-arpa: /opt/models/nnet_ms_online/G.carpa #phone-syms: /opt/models/nnet_ms_online/phones.txt #word-boundary-file: /opt/models/nnet_ms_online/word_boundary.int #do-phone-alignment: true out-dir: /opt/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);} s/(.*)/\1/;'

#A sample full post processor that add a confidence score to 1-best hyp and deletes other n-best hyps full-post-processor: /opt/kaldi-gstreamer-server/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]

Note that you must say use-nnet2: True and nnet-mode: 3.

[lukeinator42]

Just tried this and it works. you have to set frame-subsamping-factor: 3 and acoustic-scale: 1 though.