hi,
I have run all the steps from 00 to 08 in the scripts_pl. After checking the "logdir" I found few "warning and error" messages in chinna/logdir/02.ci_schmm/chinna1.norm.log
chinna/logdir/04.cd_schmm_untied/chinna1.norm.log
chinna/logdir/07.cd_schmm/chinna1.norm.log
The above files are the first iterations of "training context independent models", "training CD-untied models", and "training CD tied models" respectively.
and getting only warning messages in other iterations( i.e., from second iteration onwards).
The "warning and error" messages that I am getting are as under:
warning: "gauden.c", line 1382: (mgau=0, feat=2, density=255) never observed ..........(some more warnings for different density values)
error: "gauden.c", line 1424: var(mgau=0, feat=0, density=13, component=0)<0
..............(some more error messages for different values of density and component)
At the end of last iteration its saying that "*********TRAINING COMPLETE*********"
I have recorded 62 different sentences and have 247 words in the dictionary and 40 phones in the chinna.phone file.
Can anyone help me "why i am getting this kind of warnings and errors".
Thanx in advance.........
Sreenu.G
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2004-02-12
This type of error has been seen and documented before in the archives of this forum. Typically what is happening here is there is not enough audio data all the 6000+ tri-phones are not being represented in your training set. You need at least 8 hours of audio data to train an acoustic model, you'll see as you acquire more data you'll experience less warning messages.
Steve
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
hi,
I have run all the steps from 00 to 08 in the scripts_pl. After checking the "logdir" I found few "warning and error" messages in chinna/logdir/02.ci_schmm/chinna1.norm.log
chinna/logdir/04.cd_schmm_untied/chinna1.norm.log
chinna/logdir/07.cd_schmm/chinna1.norm.log
The above files are the first iterations of "training context independent models", "training CD-untied models", and "training CD tied models" respectively.
and getting only warning messages in other iterations( i.e., from second iteration onwards).
The "warning and error" messages that I am getting are as under:
warning: "gauden.c", line 1382: (mgau=0, feat=2, density=255) never observed ..........(some more warnings for different density values)
error: "gauden.c", line 1424: var(mgau=0, feat=0, density=13, component=0)<0
..............(some more error messages for different values of density and component)
At the end of last iteration its saying that "*********TRAINING COMPLETE*********"
I have recorded 62 different sentences and have 247 words in the dictionary and 40 phones in the chinna.phone file.
Can anyone help me "why i am getting this kind of warnings and errors".
Thanx in advance.........
Sreenu.G
This type of error has been seen and documented before in the archives of this forum. Typically what is happening here is there is not enough audio data all the 6000+ tri-phones are not being represented in your training set. You need at least 8 hours of audio data to train an acoustic model, you'll see as you acquire more data you'll experience less warning messages.
Steve