Error in PLDA scoring

[Priyanshu Soni]

Hello,

I fine-tuned my embedding extractor for speaker diarization on 100 hours of speaker verification data using the script linked below, the finetuning process ran without any bugs but when I try to compute a DER by utilizing the finetuned model for embedding extractor it is leading to errors during PLDA Scoring.

Keep in mind the recipe used to run fine before I fine-tuned.  What could be the possible issue?

I have already verified and the issue is not with the data extraction process. To the best of my knowledge it is going wrong during xvector extraction and subsequently PLDA scoring.

run_xvector_1a.sh

Error log generated during PLDA :

# Running on compute-0-3.local

# Started at Wed Jul 17 10:04:52 IST 2024

# ivector-plda-scoring-dense --target-energy=0.3 exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_0.25/plda ark:exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_scores/tmp/split4/1/spk2utt "ark:ivector-subtract-global-mean exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_0.25/mean.vec scp:exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_scores/tmp/split4/1/feats.scp ark:- | transform-vec exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_0.25/transform.mat ark:- ark:- | ivector-normalize-length ark:- ark:- |" ark,scp:exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_scores/scores.1.ark,exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_scores/scores.1.scp

ivector-plda-scoring-dense --target-energy=0.3 exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_0.25/plda ark:exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_scores/tmp/split4/1/spk2utt 'ark:ivector-subtract-global-mean exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_0.25/mean.vec scp:exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_scores/tmp/split4/1/feats.scp ark:- | transform-vec exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_0.25/transform.mat ark:- ark:- | ivector-normalize-length ark:- ark:- |' ark,scp:exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_scores/scores.1.ark,exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_scores/scores.1.scp

transform-vec exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_0.25/transform.mat ark:- ark:-

ivector-subtract-global-mean exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_0.25/mean.vec scp:exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_scores/tmp/split4/1/feats.scp ark:-

ivector-normalize-length ark:- ark:-

ASSERTION_FAILED (ivector-subtract-global-mean[5.5.1060~1-ae8cb]:AddVec<float>():kaldi-vector.cc:77) Assertion failed: (dim_ == v.dim_)

[ Stack-Trace: ]

ivector-subtract-global-mean(kaldi::MessageLogger::LogMessage() const+0x98a) [0x4a9ad6]

ivector-subtract-global-mean(kaldi::KaldiAssertFailure_(char const*, char const*, int, char const*)+0x5b) [0x4aa5f8]

ivector-subtract-global-mean(void kaldi::VectorBase<float>::AddVec<float>(float, kaldi::VectorBase<float> const&)+0x4e) [0x476b5e]

ivector-subtract-global-mean(main+0x7e0) [0x42a6e7]

/lib64/libc.so.6(__libc_start_main+0xf5) [0x2ae9692e4c05]

ivector-subtract-global-mean() [0x429e59]

LOG (transform-vec[5.5.1060~1-ae8cb]:main():transform-vec.cc:85) Applied transform to 0 vectors.

LOG (ivector-normalize-length[5.5.1060~1-ae8cb]:main():ivector-normalize-length.cc:90) Processed 0 iVectors.

ERROR (ivector-plda-scoring-dense[5.5.1060~1-ae8cb]:main():ivector-plda-scoring-dense.cc:173) No iVector present in input for utterance B007-0001544-0009340-00000000-00000150

[ Stack-Trace: ]

ivector-plda-scoring-dense(kaldi::MessageLogger::LogMessage() const+0x98a) [0x4b1b40]

ivector-plda-scoring-dense(kaldi::MessageLogger::LogAndThrow::operator=(kaldi::MessageLogger const&)+0x11) [0x42f6b9]

ivector-plda-scoring-dense(main+0x6cb) [0x42dfb6]

/lib64/libc.so.6(__libc_start_main+0xf5) [0x2ad74d94fc05]

ivector-plda-scoring-dense() [0x42d149]

WARNING (ivector-plda-scoring-dense[5.5.1060~1-ae8cb]:Close():kaldi-io.cc:515) Pipe ivector-subtract-global-mean exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_0.25/mean.vec scp:exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_scores/tmp/split4/1/feats.scp ark:- | transform-vec exp_pyannote_final/displace2024_diarization_nnet_1a_dev_fbank_spectral_laplacian_scaling10/plda_0.25/transform.mat ark:- ark:- | ivector-normalize-length ark:- ark:- | had nonzero return status 256

kaldi::KaldiFatalError# Accounting: time=1 threads=1

# Finished at Wed Jul 17 10:04:53 IST 2024 with status 255