You actual problem is that you have only 198 mixtures but want to have 4096 rows in codebook. Make more mixtures or use smaller amount of raws with -svqrows.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
We don't know too. There is just not enough information to solve this, we have no idea what you are doing, why do you run subvq, how do you run it and so on.
> Thanks !!
No problem
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
As I've read, to create the file to be specified to the flag "subvq" during the execution of the livepretend command, I have to execute the sphinx3_gausubvq command. A list of parameters must be specified to this command, so I wrote the following "gausubvq_script":
As we can see, I just set values to the flags which need file paths as input, and one of the 2 possible values for the flag "svspec". The rest of the flags, as I've read in the documentation, are supposed to have default values, so I don't set them.
Once I did this, I execute sphinx3_gausubvq as follows:
perl sphinx3_gausubvq ../../../subvq_config
And I get all this:
INFO: info.c(65): Host: 'tes-150'
INFO: info.c(66): Directory: '/home/xavi/sphinxnist_phones/sphinx3/src/programs'
INFO: info.c(70): /home/xavi/sphinxnist_phones/sphinx3/src/programs/.libs/lt-sphinx3_gausubvq Compiled on: Mar 10 2008, AT: 10:33:25
Hi,
I get this error during sphinx3_gausubvq execution:
lt-sphinx3_gausubvq: vector.c:393: vector_vqgen: Assertion `(rows >= vqrows) && (maxiter >= 0) && (epsilon > 0.0)' failed.
Cancelled (core dumped)
I don't know how to fix it...
Thanks !!
Could anyone send me some example about how you execute "sphinx3_gausubvq", please ??
I still have the same problem with it, so I guess it's because I don't give to the command the correct information...
Thanks !!!
Try sphinx3/src/tests/regression/test-gausubvq.sh
You actual problem is that you have only 198 mixtures but want to have 4096 rows in codebook. Make more mixtures or use smaller amount of raws with -svqrows.
> I don't know how to fix it...
We don't know too. There is just not enough information to solve this, we have no idea what you are doing, why do you run subvq, how do you run it and so on.
> Thanks !!
No problem
Hi,
As I've read, to create the file to be specified to the flag "subvq" during the execution of the livepretend command, I have to execute the sphinx3_gausubvq command. A list of parameters must be specified to this command, so I wrote the following "gausubvq_script":
-mean /home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/means
-mixw /home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/mixture_weights
-subvq /home/xavi/sphinxnist_phones/subvq_esentencias
-svspec 0-12/13-25/26-38
-var /home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/variances
As we can see, I just set values to the flags which need file paths as input, and one of the 2 possible values for the flag "svspec". The rest of the flags, as I've read in the documentation, are supposed to have default values, so I don't set them.
Once I did this, I execute sphinx3_gausubvq as follows:
perl sphinx3_gausubvq ../../../subvq_config
And I get all this:
INFO: info.c(65): Host: 'tes-150'
INFO: info.c(66): Directory: '/home/xavi/sphinxnist_phones/sphinx3/src/programs'
INFO: info.c(70): /home/xavi/sphinxnist_phones/sphinx3/src/programs/.libs/lt-sphinx3_gausubvq Compiled on: Mar 10 2008, AT: 10:33:25
INFO: cmd_ln.c(265): Parsing command lines from file ../../../subvq_config
INFO: cmd_ln.c(444): Parsing command line:
\ -mean /home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/means \ -mixw /home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/mixture_weights \ -subvq /home/xavi/sphinxnist_phones/subvq_esentencias \ -svspec 0-12/13-25/26-38 \ -var /home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/variances
Current configuration:
[NAME] [DEFLT] [VALUE]
-eps 0.0001 1.000000e-04
-iter 100 100
-log3table yes yes
-logbase 1.0003 1.000300e+00
-logfn
-mean /home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/means
-mixw /home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/mixture_weights
-mixwfloor 0.0000001 1.000000e-07
-seed -1 -1
-stdev no no
-subvq /home/xavi/sphinxnist_phones/subvq_esentencias
-svqrows 4096 4096
-svspec 0-12/13-25/26-38
-var /home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/variances
-varfloor 0.0001 1.000000e-04
INFO: Initialization of the log add table
INFO: Log-Add table size = 29350 x 2 >> 0
INFO:
INFO: cont_mgau.c(161): Reading mixture gaussian file '/home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/means'
INFO: cont_mgau.c(417): 189 mixture Gaussians, 8 components, 1 streams, veclen 39
INFO: cont_mgau.c(161): Reading mixture gaussian file '/home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/variances'
INFO: cont_mgau.c(417): 189 mixture Gaussians, 8 components, 1 streams, veclen 39
INFO: cont_mgau.c(505): Reading mixture weights file '/home/xavi/sphinxnist_phones/esentencias/model_parameters/esentencias.cd_cont_1000/mixture_weights'
INFO: cont_mgau.c(657): Read 189 x 8 mixture weights
INFO: cont_mgau.c(685): Removing uninitialized Gaussian densities
WARNING: "cont_mgau.c", line 760: 92 densities removed (0 mixtures removed entirely)
INFO: main_gausubvq.c(282): Clustering subvector 0
INFO: main_gausubvq.c(318): Sanity check: input data[0]:
-1.2037e+00 6.1163e-01 -1.5678e-01 3.5478e-02 8.4194e-02 1.4669e-02 -9.0781e-02 1.8168e-02 2.3995e-02 1.6741e-02 5.1470e-02 1.2079e-02 -2.6878e-02 1.3042e-02 -8.3344e-02 1.5915e-02 1.0252e-01 1.0038e-02 1.1335e-01 7.0373e-03 1.1155e-02 5.5393e-03 4.2836e-03 5.7015e-03 5.1161e-02 5.7827e-03
lt-sphinx3_gausubvq: vector.c:393: vector_vqgen: Assertion `(rows >= vqrows) && (maxiter >= 0) && (epsilon > 0.0)' failed.
Cancelado (core dumped)