plot accuracy train and valid.

[Alaa Jaber]

When run "steps/nnet3/report/generate_plots.py ", I get this result

I can not found plot for accuracy.

How can generate it?

My "compute_prob_valid.*.log" and "compute_prob_train.*.log" does not content any attribute for accuracy !!

[Daniel Povey]

The chain/LF-MMI training doesn't compute accuracy as a diagnostic.

--
Go to http://kaldi-asr.org/forums.html find out how to join
---
You received this message because you are subscribed to the Google Groups "kaldi-help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kaldi-help+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/kaldi-help/fca25921-2f59-4de4-acbd-7a682e57b745%40googlegroups.com.

[mahmoud h Hammouri]

Hi Dan ... pls what is the way to compute that accuracy ,,, and ther is any way to display or plotting the training and loss accuracy in Kaldi as this (in keras) ?????

Thanks..

بتاريخ الثلاثاء، 22 أكتوبر، 2019 7:19:52 م UTC+3، كتب Dan Povey:

The chain/LF-MMI training doesn't compute accuracy as a diagnostic.

On Tue, Oct 22, 2019 at 9:12 AM Alaa Jaber <alaa.s...@gmail.com> wrote:

When run "steps/nnet3/report/generate_plots.py ", I get this result

I can not found plot for accuracy.

How can generate it?

My "compute_prob_valid.*.log" and "compute_prob_train.*.log" does not content any attribute for accuracy !!

--
Go to http://kaldi-asr.org/forums.html find out how to join
---
You received this message because you are subscribed to the Google Groups "kaldi-help" group.

To unsubscribe from this group and stop receiving emails from it, send an email to kaldi...@googlegroups.com.

[Daniel Povey]

No; it's a sequence objective and  it isn't clear what accuracy measure is relevant.  You could manually compute the WER at various iterations if you wanted.

To unsubscribe from this group and stop receiving emails from it, send an email to kaldi-help+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/kaldi-help/e384ff5d-70b8-4727-8cfc-70af54400526%40googlegroups.com.

[mahmoud h Hammouri]

Can you tell me, from where I can get WER for each Iteration?

How Can detect if there over-fitting in model or not?

[Jan Trmal]

The decoding scripts have often --iter parameter which can be used to decide using a specific iteration. Running it by default would be computationally too expensive.

Y.

On Thu, Oct 24, 2019 at 03:03 mahmoud h Hammouri <mahmoudha...@gmail.com> wrote:

Can you tell me, from where I can get WER for each Iteration?

How Can detect if there over-fitting in model or not?

--

Go to http://kaldi-asr.org/forums.html find out how to join
---
You received this message because you are subscribed to the Google Groups "kaldi-help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kaldi-help+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/kaldi-help/c339f070-da48-4d20-9d57-52ca0443ed46%40googlegroups.com.

[Alaa Jaber]

Thank you Dan & Yenda

but my main question How can detect if there over-fitting in my model or not?