repeted entries

[hariv...@gmail.com]

A properly validated data folder using validate_data_dir.sh ......generates repeated entries after  feature extraction in ark, scp  files why does this happen?

[Daniel Povey]

This question is too non-specific to answer.

> --
> 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.
> For more options, visit https://groups.google.com/d/optout.

[Harikrishna Vydana]

i have generated tri3b,tri4 systems with a data now when i run run_bnf.sh i gives error :
ERROR (transform-feats:FindKeyInternal():util/kaldi-table-inl.h:2086) Error in RandomAccessTableReader: duplicate key mjln0 in archive cat exp/tri3b/decode_tg_test/trans.*

if there is a duplicate entry then tri3b and tri4 will show the same error? i have validated the data directories

i am using the --nj equal to number of jobs used while making tri3b..i could not resolve the error

You received this message because you are subscribed to a topic in the Google Groups "kaldi-help" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/kaldi-help/Yc5FlJVgOQ8/unsubscribe.
To unsubscribe from this group and all its topics, send an email to kaldi-help+...@googlegroups.com.

[Daniel Povey]

See if that key appears in more than one of the utt2spk files in the
split data directory that you are using (e.g.
data/test/split10/*/utt2spk). I imagine it does. If it does, then
it's likely because at some point you (or someone) ran
split_data_dir.sh with the --per-utt option. That is only called from
a small number of splits. The solution is then to remove the split
directory and rerun the decoding.

If it does not appear in more than one of the utt2spk files, then
likely the problem is that you previously ran the decoding with a
different number of jobs, leaving higher-numbered .trans files there,
and the solution is to clean up that directory.

Dan