[Subject Prev][Subject Next][Thread Prev][Thread Next][Date Index][Thread Index]

[hts-users:00039] Re: problems with HTS learning


Sorry for long delay.

I found solution while dividing set of files used for trainning. Total count of files was 1421 and beginning form file 1302 Hinit while calculating generates nan while calculating. I looked at file 1165 and found that in this file sentence start much later then in other files. Initial silence is not ~0.2 s like in other files but much longer. I've also found that in other files between 1421 and 1302 are similiar problems. On this files also SphinxTrain have problems. So i think that it must be some problem with long pau while calculating HMM.


Iin attachment You will find files. I'm not sure that error will take place if You will try this files without earlier files.



I modified questions file, but i have error on new, just created questions names for new phones i.e. :

QS "LL-YVowel"                {y^*}

What kind of error happened ?
Please send error message.


I don't know what problem it was- i've made some changes to question file and now it works.

On my computer Athlon 1.8 Ghz and 512 RAM HHed uses too much RAM (1 GB Swap and all RAM) and calculates *very* slow since 7 days.

I will try move this job to supercomputer with Intel Itanium II procesors and 2 GB RAM in one node. But i'm affraid of little and big endian problems.

best regards

Lukasz Osowski

Attachment: ivo_pl_ivona_jl_1302.cmp
Description: Binary data

         0    7900000 pau
   7900000    8500000 t
   8500000    9000000 y
   9000000   10300000 si
  10300000   11200000 o
  11200000   11700000 n
  11700000   12900000 c
  12900000   14000000 dzi
  14000000   14500000 e
  14500000   14900000 w
  14900000   15200000 j
  15200000   16000000 e
  16000000   16600000 ni
  16600000   16900000 ci
  16900000   17900000 s
  17900000   18700000 e
  18700000   19500000 t
  19500000   20800000 si
  20800000   21100000 e
  21100000   21700000 d
  21700000   22200000 e
  22200000   22800000 m
  22800000   23600000 dzi
  23600000   24100000 e
  24100000   25300000 si
  25300000   25900000 o
  25900000   26300000 n
  26300000   26800000 t
  26800000   28600000 dzi
  28600000   29300000 e
  29300000   29700000 w
  29700000   30300000 j
  30300000   31000000 e
  31000000   31800000 ni
  31800000   34600000 ci
  34600000   40300000 pau

References
[hts-users:00025] problems with HTS learning, Łukasz Osowski
[hts-users:00026] Re: problems with HTS learning, Heiga Zen / Byung-Ha Chun
[hts-users:00027] Re: problems with HTS learning, Łukasz Osowski
[hts-users:00028] Re: problems with HTS learning, Heiga Zen / Byung-Ha Chun
[hts-users:00030] Re: problems with HTS learning, Łukasz Osowski
[hts-users:00031] Re: problems with HTS learning, Heiga Zen / Byung-Ha Chun
[hts-users:00033] Re: problems with HTS learning, Łukasz Osowski
[hts-users:00034] Re: problems with HTS learning, Heiga Zen / Byung-Ha Chun
[hts-users:00037] Re: problems with HTS learning, Łukasz Osowski
[hts-users:00038] Re: problems with HTS learning, Heiga Zen / Byung-Ha Chun