Decode script does not generate 1Best.ctm

[akashrajkn]

Hi, thank you for open sourcing this project.

I am trying to run the nnet3_online tdnn model. On running the decode.sh script, I get the following output

Argument filename.wav is a sound file, using it as audio
Diarization completed in 0:00:47 (CPU: 0:00:40), Memory used: 117 MB                
Split 1 source file into 103 segments                              
Duration of speech: 0h:10m:23s
NNet3 decoding completed in 0:00:00 (CPU: 0:00:00), Memory used: 4 MB                
Rescoring completed in 0:00:00 (CPU: 0:00:00), Memory used: 0 MB                
Rescoring completed in 0:00:00 (CPU: 0:00:00), Memory used: 2 MB                
Done

However does not contain 1Best.ctm results file. Any help is appreciated.

[akashrajkn]

On investigation of output/intermediate/log file, I see that the following files are missing:

models/NL/UTwente/HMI/AM/CGN_all/nnet3_online/tdnn/v1.0/graph_UTwente_HMI_KrantenTT_v1.0_KrantenTT.3gpr.kn.int_UTwente_HMI_lexicon/HCLG.fst

models/NL/UTwente/HMI/LM/KrantenTT/v1.0/LG_KrantenTT.3gpr.kn.int_UTwente_HMI_lexicon/G.fst

[thatux]

same problem here, I'm using WSL, everything seems to be fine, but no 1Best.ctm

command

./decode_PR.sh /mnt/e/filtered_short.mp3 output/3 Argument /mnt/e/filtered_short.mp3 is a sound file, using it as audio [ ] Diarization (0/1) Diarization completed in 0:01:13 (CPU: 0:01:34), Memory used: 589 MB Split 1 source file into 361 segments Duration of speech: 0h:50m:0s NNet3 decoding completed in 0:00:00 (CPU: 0:00:01), Memory used: 5 MB Rescoring completed in 0:00:00 (CPU: 0:00:00), Memory used: 0 MB Rescoring completed in 0:00:00 (CPU: 0:00:00), Memory used: 1 MB Done

log

./decode_PR.sh /mnt/e/filtered_short.mp3 output/3 cat: 'output/3/intermediate/data/.stm': No such file or directory fix_data_dir.sh: kept all 361 utterances. fix_data_dir.sh: old files are kept in output/3/intermediate/data/ALL/.backup steps/make_mfcc.sh --nj 8 --mfcc-config output/3/intermediate/mfcc.conf output/3/intermediate/data/ALL output/3/intermediate/data/ALL/log output/3/intermediate/mfcc Usage: utils/validate_data_dir.sh [--no-feats] [--no-text] [--non-print] [--no-wav] [--no-spk-sort] The --no-xxx options mean that the script does not require xxx.scp to be present, but it will check it if it is present. --no-spk-sort means that the script does not require the utt2spk to be sorted by the speaker-id in addition to being sorted by utterance-id. --non-print ignore the presence of non-printable characters. By default, utt2spk is expected to be sorted by both, which can be achieved by making the speaker-id prefixes of the utterance-ids e.g.: utils/validate_data_dir.sh data/train steps/compute_cmvn_stats.sh output/3/intermediate/data/ALL output/3/intermediate/data/ALL/log output/3/intermediate/mfcc steps/compute_cmvn_stats.sh: no such file output/3/intermediate/data/ALL/feats.scp steps/online/nnet3/decode.sh --nj 8 --acwt 1.2 --post-decode-acwt 10.0 --skip-scoring true models/AM/graph_PR output/3/intermediate/data/ALL output/3/tmp/tmp.Xzb6VJ5zsx filter_scps.pl: warning: some input lines were output to multiple files [OK if splitting per utt] run.pl: 8 / 8 failed, log is in output/3/tmp/tmp.Xzb6VJ5zsx/log/decode..log steps/lmrescore_const_arpa.sh --skip-scoring true models/Lang_PR models/LM/Const_PR output/3/intermediate/data/ALL output/3/intermediate/decode output/3/intermediate/rescore ls: cannot access 'output/3/intermediate/decode/lat.*.gz': No such file or directory steps/lmrescore_const_arpa .sh: No lattices input directory output/3/intermediate/decode

[proycon]

Sorry for the long delay! I'm investigating these issues now.

[proycon]

I have just released Kaldi-NL v0.3 which should hopefully fix this problem. Please let me know if it still persists.