Menu

Mixture Gaussians doesn't match mean/var

Help
2010-05-10
2012-09-22
  • ubaid mahmood

    ubaid mahmood - 2010-05-10

    Hello,

    I am running the preliminary training run and am getting the following error.

    MODULE: 11 Force-aligning transcripts (2010-05-09 22:36)
    
    Phase 1: Cleaning up directories:
    logs... output... qmanager...
    
    Phase 3: Creating dictionary for alignment...
    
    Phase 4: Creating transcript for alignment...
    
    Phase 5: Running force alignment in 1 parts
    
    Force alignment starting: (1 of 1)
    
    sphinx3_align Log File
    
    FATAL_ERROR: "cont_mgau.c", line 582: C:/Working/sphinx/tutorial/doheny/model_parameters/doheny.falign_ci_cont/mixture_weights: #Mixture Gaussians(153) doesn't match mean/var parameters(156)
    FAILED
    
    Failed to run sphinx3_align
    

    I do not understand the error, nor how to resolve it. Any assistance would be
    helpful.

    Thanks.

     
  • ubaid mahmood

    ubaid mahmood - 2010-05-10

    Seems that the mdef file and the mean and variances files are not in sync. Is
    this a sphinx train configuration issues because it seems that sphinx train
    should be generating these files on its own.

    INFO: Reading HMM in Sphinx 3 Model format
    INFO: Model Definition File:
    C:/Working/sphinx/tutorial/doheny/model_architecture/doheny.falign_ci.mdef
    INFO: Mean File:
    C:/Working/sphinx/tutorial/doheny/model_parameters/doheny.falign_ci_cont/means
    INFO: Variance File: C:/Working/sphinx/tutorial/doheny/model_parameters/doheny
    .falign_ci_cont/variances
    INFO: Mixture Weight File: C:/Working/sphinx/tutorial/doheny/model_parameters/
    doheny.falign_ci_cont/mixture_weights
    INFO: Transition Matrices File: C:/Working/sphinx/tutorial/doheny/model_parame
    ters/doheny.falign_ci_cont/transition_matrices
    INFO: mdef.c(682): Reading model definition:
    C:/Working/sphinx/tutorial/doheny/model_architecture/doheny.falign_ci.mdef
    INFO: Initialization of mdef_t, report:
    INFO: 51 CI-phone, 0 CD-phone, 3 emitstate/phone, 153 CI-sen, 153 Sen, 51 Sen-
    Seq
    INFO:
    INFO: kbcore.c(298): Using optimized GMM computation for Continuous HMM, -topn
    will be ignored
    INFO: cont_mgau.c(163): Reading mixture gaussian file 'C:/Working/sphinx/tutor
    ial/doheny/model_parameters/doheny.falign_ci_cont/means'
    INFO: cont_mgau.c(422): 156 mixture Gaussians, 1 components, 1 streams, veclen
    39
    INFO: cont_mgau.c(163): Reading mixture gaussian file 'C:/Working/sphinx/tutor
    ial/doheny/model_parameters/doheny.falign_ci_cont/variances'
    INFO: cont_mgau.c(422): 156 mixture Gaussians, 1 components, 1 streams, veclen
    39
    INFO: cont_mgau.c(523): Reading mixture weights file 'C:/Working/sphinx/tutori
    al/doheny/model_parameters/doheny.falign_ci_cont/mixture_weights'
    FATAL_ERROR: "cont_mgau.c", line 582: C:/Working/sphinx/tutorial/doheny/model_
    parameters/doheny.falign_ci_cont/mixture_weights: #Mixture Gaussians(153)
    doesn't match mean/var parameters(156)
    Sun May 9 22:36:32 2010

     
  • Nickolay V. Shmyrev

    Is this a sphinx train configuration issues because it seems that sphinx
    train should be generating these files on its own.

    No, it isn't

    I do not understand the error, nor how to resolve it. Any assistance would
    be helpful.

    You need to check the log how models for forced alignment were built. This is
    done on previous stage 10. Most likely the models were not updated for some
    reason like permision issue.

     

Log in to post a comment.