Long audio anonymization fails
After gathering data for scientific evaluation, I noticed that audio transformation with privacy_driven_speech_transformation method was often unsuccessful. This mostly happened when audio duration exceeded ~1:30. I successfully reproduced the problem by sending ~2min audio data to comprise personal server and received error in comprise-st container:
Execute command:
/opt/nii/CURRENNT_codes/build/currennt --train false --ff_output_format htk --parallel_sequences 1 --input_noise_sigma 0 --random_seed 12345231 --shuffle_fractions false --shuffle_sequences false --revert_std true --network exp/models/4_nsf/trained_network.jsn --ff_output_file /opt/vpc/exp/am_nsf_data/single_wav/nsf_output_wav --ff_input_file /opt/vpc/exp/am_nsf_data/single_wav/output_tmp/ncData/DATA_TEST/data.nc1 --ExtInputDirs /opt/vpc/exp/am_nsf_data/single_wav/am_out_mel,/opt/vpc/exp/am_nsf_data/single_wav/xvector,/opt/vpc/exp/am_nsf_data/single_wav/f0 --ExtInputExts .mel,.xvector,.f0 --ExtInputDims 80_512_1 --resolutions 160 --waveNetMemSave 1
Failed to run:/opt/nii/CURRENNT_codes/build/currennt --train false --ff_output_format htk --parallel_sequences 1 --input_noise_sigma 0 --random_seed 12345231 --shuffle_fractions false --shuffle_sequences false --revert_std true --network exp/models/4_nsf/trained_network.jsn --ff_output_file /opt/vpc/exp/am_nsf_data/single_wav/nsf_output_wav --ff_input_file /opt/vpc/exp/am_nsf_data/single_wav/output_tmp/ncData/DATA_TEST/data.nc1 --ExtInputDirs /opt/vpc/exp/am_nsf_data/single_wav/am_out_mel,/opt/vpc/exp/am_nsf_data/single_wav/xvector,/opt/vpc/exp/am_nsf_data/single_wav/f0 --ExtInputExts .mel,.xvector,.f0 --ExtInputDims 80_512_1 --resolutions 160 --waveNetMemSave 1
Please check the printed error message
Process terminated with 134
Stage a.7: Creating new data directories corresponding to anonymization.
utils/copy_data_dir.sh: copied data from data/single_wav to data/single_wav_anon
utils/validate_data_dir.sh: WARNING: you have only one speaker. This probably a bad idea.
Search for the word 'bold' in http://kaldi-asr.org/doc/data_prep.html
for more information.
utils/validate_data_dir.sh: Successfully validated data-directory data/single_wav_anon
cp: cannot stat '/opt/vpc/exp/am_nsf_data/single_wav/nsf_output_wav/*.wav': No such file or directory
172.17.0.1 - - [10/Nov/2021 13:00:35] "POST /vpc?%7B%22wgender%22:%20%22f%22,%20%22cross_gender%22:%20%22same%22,%20%22distance%22:%20%22plda%22,%20%22proximity%22:%20%22dense%22,%20%22sample-frequency%22:%2016000%7D HTTP/1.1" 500 -