Re: [Kaldi-users] Need for real time factor

[Vassil P. <vas...@gm...>]

The online decoders are inherently constrained to run faster than real
time - otherwise audio input buffers overflow and in general the
recognizer can't keep up with the speaker. These real-time factors
were meant to allow the user to specify minimum and maximum bounds in
which the inner loop of the decoder is allowed to run, by decreasing
the beam if the upper bound is exceeded and increasing it if the
decoding become too fast. In practice, I think the current
implementation isn't perfect.

Vassil

On Thu, Oct 17, 2013 at 5:22 AM, Daniel Povey <dp...@gm...> wrote:
> I think the latter; this is probably in the online decoder.
> Dan
>
>
> On Wed, Oct 16, 2013 at 7:06 PM, E <oth...@ao...> wrote:
>> Hello,
>>
>> I'm a bit flummoxed by rt-min and rt-max factors. I thought these factors
>> are computed after recognition is complete- depending on how much time it
>> took to decode. How are then they supplied as arguments to decoder? Does the
>> decoder do more pruning if it seems that rt-max is getting exceeded?
>>
>> ------------------------------------------------------------------------------
>> October Webinars: Code for Performance
>> Free Intel webinars can help you accelerate application performance.
>> Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most
>> from
>> the latest Intel processors and coprocessors. See abstracts and register >
>> http://pubads.g.doubleclick.net/gampad/clk?id=60135031&iu=/4140/ostg.clktrk
>> _______________________________________________
>> Kaldi-users mailing list
>> Kal...@li...
>> https://lists.sourceforge.net/lists/listinfo/kaldi-users
>>
>
> ------------------------------------------------------------------------------
> October Webinars: Code for Performance
> Free Intel webinars can help you accelerate application performance.
> Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from
> the latest Intel processors and coprocessors. See abstracts and register >
> http://pubads.g.doubleclick.net/gampad/clk?id=60135031&iu=/4140/ostg.clktrk
> _______________________________________________
> Kaldi-users mailing list
> Kal...@li...
> https://lists.sourceforge.net/lists/listinfo/kaldi-users