I rerun the hts-demo-slt on centos with the configuration (MGCORDER=24 MGCLSP=1 GAMMA=3 FREQWARP=0.42). A error occured at the same place just like that in v2.1rc1. That is:
- Training process goes well;
- synthesizing with 1mix goes well;
- synthesizing with hts_engine goes well;
- failed at "Start semi-tied covariance matrices", shown as below:
Processing Data: cmu_us_arctic_slt_b0539.cmp; Label cmu_us_arctic_slt_b0539.lab
Utterance prob per frame = 2.526787e+02
Class 1 (stream=1, vsize=75)
Initial LogLikelihood = 318.299340
Iteration 1
Floored 6144 component elements
LogLikelihood = 337.625052
Iteration 2
Floored 6648 component elements
LogLikelihood = 340.607893
...
Iteration 20
Floored 6837 component elements
LogLikelihood = 343.513972
Class 2 (stream=2, vsize=1)
Initial LogLikelihood = 2.390779
Iteration 1
Floored 665 component elements
LogLikelihood = 2.360921
...
Iteration 19
Floored 665 component elements
LogLikelihood = 2.360921
Iteration 20
Error in /usr/local/HTS-2.1/bin/HERest -A -B -C /home/hyf/work/HTS-demo_CMU-ARCTIC-SLT/configs/trn.cnf -D -T 1 -S /home/hyf/work/HTS-demo_CMU-ARCTIC-SLT/data/scp/train.scp -I /home/hyf/work/HTS-demo_CMU-ARCTIC-SLT/data/labels/full.mlf -m 1 -u tmvwdmv -w 3 -t 1500 100 5000 -H /home/hyf/work/HTS-demo_CMU-ARCTIC-SLT/models/qst001/ver1/cmp/stc.mmf -N /home/hyf/work/HTS-demo_CMU-ARCTIC-SLT/models/qst001/ver1/dur/stc.mmf -M /home/hyf/work/HTS-demo_CMU-ARCTIC-SLT/models/qst001/ver1/cmp -R /home/hyf/work/HTS-demo_CMU-ARCTIC-SLT/models/qst001/ver1/dur -C /home/hyf/work/HTS-demo_CMU-ARCTIC-SLT/configs/stc.cnf -K /home/hyf/work/HTS-demo_CMU-ARCTIC-SLT/models/qst001/ver1/cmp stc -u smvdmv /home/hyf/work/HTS-demo_CMU-ARCTIC-SLT/data/lists/full.list /home/hyf/work/HTS-demo_CMU-ARCTIC-SLT/data/lists/full.list
Does anyone tell me why? thanks
-yufeng