kaldi-asr / kaldi-asr/kaldi

chime6: train_diarizer.sh: missing local/make_voxceleb1_v2.pl

Open
#4,010 6 comments 0 reactions 0 assignees View on GitHub
stale
Dominant language
Shell
Stars
15.5k
Forks
5.4k
PR merge metrics
No merged PRs in 30d

Description

In the file `kaldi/egs/chime6/s5_track2/local/train_diarizer.sh` the comments suggest to use `local/make_voxceleb1_v2.pl` when voxceleb was downloaded recently.

https://github.com/kaldi-asr/kaldi/blob/b1d809386c17a9630d786b8fb0f67ed3bc173811/egs/chime6/s5_track2/local/train_diarizer.sh#L44-L45

The problem is, that this file does not exist in `chime6`.

I found this file 3 times in this repr. One symlink and two files.
```
.../kaldi $ find -iname "make_voxceleb1_v2.pl"
./egs/dihard_2018/v2/local/make_voxceleb1_v2.pl
./egs/dihard_2018/v1/local/make_voxceleb1_v2.pl
./egs/voxceleb/v1/local/make_voxceleb1_v2.pl
.../kaldi $ ll ./egs/dihard_2018/v2/local/make_voxceleb1_v2.pl
lrwxrwxrwx 1 cbj cbj 35 Jan 28 14:42 ./egs/dihard_2018/v2/local/make_voxceleb1_v2.pl -> ../../v1/local/make_voxceleb1_v2.pl*
.../kaldi $ ll ./egs/dihard_2018/v1/local/make_voxceleb1_v2.pl
-rwxrwxr-x 1 cbj cbj 4,5K Jan 28 14:42 ./egs/dihard_2018/v1/local/make_voxceleb1_v2.pl*
.../kaldi $ ll ./egs/voxceleb/v1/local/make_voxceleb1_v2.pl
-rwxrwxr-x 1 cbj cbj 4,6K Jan 28 14:42 ./egs/voxceleb/v1/local/make_voxceleb1_v2.pl*
```
After comparing the files, I think the files in `dihard_2018` have a bug, because `voxceleb` has one bugfix commit that was not applied to `dihard_2018`.

Also the other `make_voxceleb*` files are copies. Wouldn't it be better to use symlinks?

I am not sure, who is responsible for this, maybe @aarora8, @desh2608 or @sw005320.

Contributor guide

No contributing guide indexed for this repository

Research direction

Start with egs/chime6/s5_track2/local/train_diarizer.sh and inspect the referenced local/make_voxceleb1_v2.pl path. Compare it with egs/dihard_2018/v1/local/make_voxceleb1_v2.pl, egs/dihard_2018/v2/local/make_voxceleb1_v2.pl, and egs/voxceleb/v1/local/make_voxceleb1_v2.pl; done means the CHiME6 workflow's reference resolves to the intended, bug-fixed script.

Written by the indexing model from the issue text.

Assessment

Tech stack
perl, shell
Domain
machine-learning
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.