[hts-users:02439] Re: AW: Re: AW: Re: AW: HMGenS with GV
- Subject: [hts-users:02439] Re: AW: Re: AW: Re: AW: HMGenS with GV
- From: Keiichiro Oura <uratec@xxxxxxxxxxxxxxx>
- Date: Wed, 14 Apr 2010 03:34:17 +0900
- Cc: uratec <uratec@xxxxxxxxxxxxxxx>
- Delivered-to: hts-users@xxxxxxxxxxxxxxx
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:sender:received:in-reply-to :references:date:x-google-sender-auth:received:message-id:subject :from:to:cc:content-type:content-transfer-encoding; bh=FhI88b6kCs21FpPSV8dhJTKL5TTKr6WSWzNfpHlcI6Q=; b=GlHLpkA13w6HHk7vC9Y+9i4t1Nj84u8GU7kGtarwDljTSg6ZmfUHehZ7Fhxq67/thg Lrjb9X1cwKDe6Ee4MDJdBRjWHKi/HXWHYXdRkT+99A2BMWNfrifpXQtF8qlP/6Bz9GTi +USuxVAE86TP/CERBcL+ERtt0rurJ0yQra2SA=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:sender:in-reply-to:references:date :x-google-sender-auth:message-id:subject:from:to:cc:content-type :content-transfer-encoding; b=Ntt7g/h5s8RJfMWOBLJ+REc9/fiDYvqvKCR2Ib4NyAhcfhjDA7/BF78ql1Ww6qqRVZ v7h6Uv5ULR4hZyX7zenQArrT/IuY86AUbNeftanCZCciZpuczwHEQb37M31sbOpcr6Mb DEr9NHj6JWsjM1EK5DrWzvoV22LFPtWbhqLtY=
Hi, Florian
I found the bug in HTKLib/HGen.c (HTS-2.1.1 beta).
I would like to fix it in HTS-2.1.1.
Thank you for your report.
Regards,
Keiichiro Oura
2010/4/13 Florian Krebs <flokadillo@xxxxxxxx>:
> Thanks for your answer, Keiichiro. I was off for some days, that's why my
> answer took so long.
>
> So "HMGenS -s" is working with me with the following two label formats:
>
> 0 1000 model_name[2] model_name
> 1000 2000 model_name[3]
> 2000 3000 model_name[4]
> 3000 4000 model_name[5]
> 4000 5000 model_name[6]
> ...
>
> or
>
> 0 1000 .state[2] model_name
> 1000 2000 .state[3] model_name
> 2000 3000 .state[4] model_name
> 3000 4000 .state[5] model_name
> 4000 5000 .state[6] model_name
> ...
>
> HMGenS doesn't work with GV-HSMM models I guess, that's why you can't turn
> GV on.
>
> Best greetings,
>
> Florian Krebs
>
>
>
>
> ________________________________
> Von: Keiichiro Oura <uratec@xxxxxxxxxxxxxxx>
> An: hts-users@xxxxxxxxxxxxxxx
> Gesendet: Donnerstag, den 1. April 2010, 16:42:41 Uhr
> Betreff: [hts-users:02429] Re: AW: Re: AW: HMGenS with GV
>
> Hi,
>
> Could you try to use the labels with the following format in HMGenS?
>
> 0 1000 model_name[2] model_name
> 1000 2000 model_name[3]
> 2000 3000 model_name[4]
> 3000 4000 model_name[5]
> 4000 5000 model_name[6]
> ...
>
> Regards,
> Keiichiro Oura
>
> 2010/3/30 Florian Krebs <flokadillo@xxxxxxxx>:
>> Hello Keiichiro,
>>
>> Thank you for your answer. Are you sure the label files (that are used as
>> input for HMGenS) have to be this format ?
>>
>> When I use the label file produced by HVite with the following format:
>>
>> 0 50000
>>
>> x^x-pau+hh=aa@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:1+1+4/D:0_0/E:x+x@x+x&x+x#x+x/F:content_1/G:0_0/H:x=x@1=2|0/I:10=7/J:17+12-2[2]
>> 50000 100000
>>
>> x^x-pau+hh=aa@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:1+1+4/D:0_0/E:x+x@x+x&x+x#x+x/F:content_1/G:0_0/H:x=x@1=2|0/I:10=7/J:17+12-2[3]
>> 100000 150000
>>
>> x^x-pau+hh=aa@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:1+1+4/D:0_0/E:x+x@x+x&x+x#x+x/F:content_1/G:0_0/H:x=x@1=2|0/I:10=7/J:17+12-2[4]
>>
>> I get the following error:
>>
>> ERROR [+9999] InitGenInfo: Invalid label format for parameter
>> generation with state alignments
>>
>> I changed the label format because of Hui LIANGs posting
>> [hts-users:01991].
>> With disabling GV HMGenS runs without error message, but the result sounds
>> terrible.
>>
>> Greetings,
>>
>> Florian Krebs
>>
>> ________________________________
>> Von: Keiichiro Oura <uratec@xxxxxxxxxxxxxxx>
>> An: hts-users <hts-users@xxxxxxxxxxxxxxx>
>> Gesendet: Montag, den 29. März 2010, 16:19:26 Uhr
>> Betreff: [hts-users:02424] Re: AW: HMGenS with GV
>>
>> Hi,
>>
>> You may fail to generate labels.
>> This is an example of correct state alignment labels.
>> Please, check your labels.
>>
>> 0 4200000
>>
>> x^x-pau+pau=ao@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:0+0+0/D:0_0/E:x+x@x+x&x+x#x+x/F:0_0/G:0_0/H:x=x@1=2|0/I:0=0/J:14+8-2[2]
>>
>> x^x-pau+pau=ao@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:0+0+0/D:0_0/E:x+x@x+x&x+x#x+x/F:0_0/G:0_0/H:x=x@1=2|0/I:0=0/J:14+8-2
>> 4200000 9250000
>>
>> x^x-pau+pau=ao@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:0+0+0/D:0_0/E:x+x@x+x&x+x#x+x/F:0_0/G:0_0/H:x=x@1=2|0/I:0=0/J:14+8-2[3]
>> 9250000 14000000
>>
>> x^x-pau+pau=ao@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:0+0+0/D:0_0/E:x+x@x+x&x+x#x+x/F:0_0/G:0_0/H:x=x@1=2|0/I:0=0/J:14+8-2[4]
>> 14000000 14050000
>>
>> x^x-pau+pau=ao@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:0+0+0/D:0_0/E:x+x@x+x&x+x#x+x/F:0_0/G:0_0/H:x=x@1=2|0/I:0=0/J:14+8-2[5]
>> 14050000 14100000
>>
>> x^x-pau+pau=ao@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:0+0+0/D:0_0/E:x+x@x+x&x+x#x+x/F:0_0/G:0_0/H:x=x@1=2|0/I:0=0/J:14+8-2[6]
>> ...
>>
>> Regards,
>> Keiichiro Oura
>>
>> 2010/3/29 Florian Krebs <flokadillo@xxxxxxxx>
>>>
>>> I switched GV off (USEGV = FALSE in syn.cnf) and now it is working.
>>> Greetings,
>>>
>>> Florian Krebs
>>>
>>> ________________________________
>>> Von: Florian Krebs <flokadillo@xxxxxxxx>
>>> An: hts-users@xxxxxxxxxxxxxxx
>>> Gesendet: Montag, den 29. März 2010, 15:37:58 Uhr
>>> Betreff: [hts-users:02422] HMGenS with GV
>>>
>>> Hello,
>>>
>>> I want to use HMGenS with the -s option (use state alignment for
>>> duration). I did state alignment with HVite according to the mailing list
>>> [hts-users:01991] and brought my lab files to the format:
>>>
>>> 0 50000 .state[2]
>>>
>>> x^x-pau+pau=ae@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:0+0+0/D:0_0/E:
>>> [...]
>>> 50000 100000 .state[3]
>>>
>>> x^x-pau+pau=ae@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:0+0+0/D:0_0/E:
>>> [...]
>>>
>>>
>>> When I execute HMGenS with
>>>
>>> HMGenS -A -B -s -C configs/syn.cnf -D -T 1 -S
>>> data/scp/aligned_gen.scp
>>> -t 1500 100 5000 -c 0 \
>>> -H
>>>
>>> models/qst001/ver1/cmp/re_clustered_all.mmf.1mix -M aligned/gen
>>> models/qst001/ver1/cmp/tiedlist
>>>
>>> I get the error:
>>>
>>> ERROR [+9935] Generator: Cannot
>>> find GV model .state[2] in current list
>>>
>>> I trained my models using the HTS-demo_CMU-ARCTIC-SLT_STRAIGHT_2.1.1_beta
>>> with GV turned on.
>>>
>>> $useGV = 1;
>>>
>>> How can I solve this problem ? Is there something wrong using the Global
>>> Variance in the parameter generation algorithm ?
>>>
>>>
>>> Thank you very much,
>>> Best greetings!
>>>
>>> Florian Krebs
>>>
>>>
>>>
>>>
>>> __________________________________________________
>>> Do You Yahoo!?
>>> Sie sind Spam leid? Yahoo! Mail verfügt über einen herausragenden Schutz
>>> gegen Massenmails.
>>> http://mail.yahoo.com
>>> __________________________________________________
>>> Do You Yahoo!?
>>> Sie sind Spam leid? Yahoo! Mail verfügt über einen herausragenden Schutz
>>> gegen Massenmails.
>>> http://mail.yahoo.com
>>
>> __________________________________________________
>> Do You Yahoo!?
>> Sie sind Spam leid? Yahoo! Mail verfügt über einen herausragenden Schutz
>> gegen Massenmails.
>> http://mail.yahoo.com
>
>
> __________________________________________________
> Do You Yahoo!?
> Sie sind Spam leid? Yahoo! Mail verfügt über einen herausragenden Schutz
> gegen Massenmails.
> http://mail.yahoo.com
- References
-
- [hts-users:02422] HMGenS with GV, Florian Krebs
- [hts-users:02423] AW: HMGenS with GV, Florian Krebs
- [hts-users:02424] Re: AW: HMGenS with GV, Keiichiro Oura
- [hts-users:02425] AW: Re: AW: HMGenS with GV, Florian Krebs
- [hts-users:02429] Re: AW: Re: AW: HMGenS with GV, Keiichiro Oura
- [hts-users:02437] AW: Re: AW: Re: AW: HMGenS with GV, Florian Krebs