Set enough memory so rebuilding repository is not triggered about alex

[oplatek]

I updated the memory but decoding for some utterances is still slow (others are fine RTF < 0.5):

Wav file:  /ha/projects/vystadial/data/call-logs/2013-05-30-alex-aotb-prototype/part4/2013-12-05-15-13-39.890152-CET-6280e847c162a9c1/vad-2013-12-05-15-14-45.876037.wav

Reference: _NOISE_ _LAUGH_ I P PAVLOVA
Decoded:   _NOISE_ I P PAVLOVA
Wav dur:   3.97
Dec dur:   32.79

NBest list:
1.000 _NOISE_ I P PAVLOVA

# waws:              400
Global RTF mean:     6.695084
RTF median:          0.315860  RTF   < 5.726351 [in 95%]
Delay median:        0.000000  Delay < 13.990000 [in 95%]

Wav file: /ha/projects/vystadial/data/call-logs/2013-05-30-alex-aotb-prototype/part4/2013-12-12-17-25-10.034704-CET-44ff0981f68a03a5/vad-2013-12-12-17-26-20.849911.w
av

Reference: _LAUGH_ JAKÉ BUDE POČASÍ V PLZNI
Decoded:   JAKÉ BUDE POČASÍ V PRAZE
Wav dur:   4.14
Dec dur:   136.96

NBest list:
1.000 JAKÉ BUDE POČASÍ V PRAZE

# waws:              387
Global RTF mean:     6.723960
RTF median:          0.312500  RTF   < 5.208333 [in 95%]
Delay median:        0.000000  Delay < 10.756000 [in 95%]

[oplatek]

Done and also now svn trunk version should prevent this

[etienne1985]

Hi,
I'm getting a similar issue... I was getting slowdown, "Rebuilding repository" and dropped words. By increasing the mem-max like you recommended I got rid of the "Rebuilding repository", but it's still slow and more significantly, stretches of words get dropped around those points. This is with straight Kaldi, no python. Any ideas what's causing this? Thank you in advance!

[oplatek]

Hi @etienne1985 ,
Can you be more specific for "stretches of words get dropped around those points"?
For clarification, are you using the onl-rec code or just plain Kaldi?