Conversation
There was a problem hiding this comment.
Pull Request Overview
Adds a complete recipe for spoken language identification on the VoxLingua107 dataset, including dataset download/preparation, training/inference orchestration, and evaluation utilities.
- Introduces data download and preparation scripts to generate Kaldi-style mappings.
- Provides a run script (
run.sh) and baseline training configuration (mms_ecapa_baseline.yaml). - Supplies scoring and data-copy utilities adapted for language identification.
Reviewed Changes
Copilot reviewed 18 out of 19 changed files in this pull request and generated 3 comments.
Show a summary per file
| File | Description |
|---|---|
| egs2/voxlingua107/lid1/utils | Symlink to shared utility scripts from TEMPLATE |
| egs2/voxlingua107/lid1/steps | Symlink to shared Kaldi-like steps from TEMPLATE |
| egs2/voxlingua107/lid1/scripts | Symlink to shared helper scripts from TEMPLATE |
| egs2/voxlingua107/lid1/run.sh | Main entry-point for training and inference |
| egs2/voxlingua107/lid1/pyscripts | Symlink to shared Python helper scripts from TEMPLATE |
| egs2/voxlingua107/lid1/path.sh | Environment path setup |
| egs2/voxlingua107/lid1/local/score.py | Computes overall, per-language, and error-frequency scores |
| egs2/voxlingua107/lid1/local/prepare_voxlingua107.py | Generates wav.scp and utt2lang with ISO3 language codes |
| egs2/voxlingua107/lid1/local/data.sh | Automates dataset download, extraction, and data preparation |
| egs2/voxlingua107/lid1/local/copy_data_dir.sh | Copies and prefixes data directories for language IDs |
| egs2/voxlingua107/lid1/lid.sh | Symlink to core recipe script from TEMPLATE |
| egs2/voxlingua107/lid1/db.sh | Database path configuration |
| egs2/voxlingua107/lid1/conf/slurm.conf | Slurm scheduler configuration |
| egs2/voxlingua107/lid1/conf/queue.conf | SGE scheduler configuration |
| egs2/voxlingua107/lid1/conf/pbs.conf | PBS scheduler configuration |
| egs2/voxlingua107/lid1/conf/mms_ecapa_baseline.yaml | ECAPA-TDNN baseline model configuration |
| egs2/voxlingua107/lid1/cmd.sh | Dispatch script for run.pl / queue.pl / slurm.pl, etc. |
| egs2/voxlingua107/lid1/README.md | Recipe overview and reported results |
Comments suppressed due to low confidence (1)
egs2/voxlingua107/lid1/README.md:11
- [nitpick] There is a stray Markdown bold marker (
**) on this line which likely causes a formatting glitch. Consider removing it.
**
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #6174 +/- ##
==========================================
+ Coverage 53.53% 55.82% +2.29%
==========================================
Files 888 889 +1
Lines 84131 84275 +144
==========================================
+ Hits 45039 47049 +2010
+ Misses 39092 37226 -1866
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
for more information, see https://pre-commit.ci
for more information, see https://pre-commit.ci
|
@ftshijt, can you review this PR? |
ftshijt
left a comment
There was a problem hiding this comment.
Thanks for sharing the update. Please remember to add the data entry in "egs2/README.md"
Thanks for your review. Already added the voxlingua107 data entry to |
|
Thanks! |
What did you change?
local/data.sh,local/prepare_voxlingua107.py: download and prepare VoxLingua107 dataset.run.sh,mms_ecapa_baseline.yaml: run script and training configuration.Why did you make this change?
This PR adds a recipe for spoken language identification (LID) on the VoxLingua107 dataset.
Is your PR small enough?
Yes.
Additional Context