Thanks to visit codestin.com
Credit goes to github.com

Skip to content

SPK recipe for CN-Celeb#6126

Merged
sw005320 merged 3 commits intoespnet:masterfrom
holvan:cnceleb
Jun 12, 2025
Merged

SPK recipe for CN-Celeb#6126
sw005320 merged 3 commits intoespnet:masterfrom
holvan:cnceleb

Conversation

@holvan
Copy link
Contributor

@holvan holvan commented Jun 2, 2025

What?

Create cnceleb/spk1 recipe based on espnet2 for speaker identification on CN-Celeb dataset.

Why?

ESPnet bootcamp project.

Copy link
Contributor

@sw005320 sw005320 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks good to me.
Currently, our CI is broken, and we cannot do further check (apologize for it...)

@Jungjee, can you also take a look at this PR?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It has a large update.
I think you changed the position of the delimiter?
It is no problem, but please make sure that you do not include any errors.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, I added a line and the delimiter was automatically updated. There is no error, but I think I can revert the unnecessary changes.


| Model (conf name) | EER(%) | minDCF | Note | Huggingface |
|---|---|---|---|---|
| [conf/train_ECAPA_mel.yaml](conf/train_ECAPA_mel.yaml) | 8.372 | 0.4552 | | https://huggingface.co/espnet/cnceleb_ecapa_mel |
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this performance SOTA?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No, actually this is the baseline result for the CN-Celeb dataset using the ECAPA-TDNN configuration in the ESPnet SPK recipe. If necessary, I can tune the configurations and update the models with better performance later.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

How much does it differ?
Yes, can you tune a model a little bit?
It would be a good practice.

@codecov
Copy link

codecov bot commented Jun 3, 2025

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 14.47%. Comparing base (3856d99) to head (6862960).
Report is 5 commits behind head on master.

❗ There is a different number of reports uploaded between BASE (3856d99) and HEAD (6862960). Click for more details.

HEAD has 7 uploads less than BASE
Flag BASE (3856d99) HEAD (6862960)
test_integration_espnetez 2 0
test_integration_espnet2 5 0
Additional details and impacted files
@@             Coverage Diff             @@
##           master    #6126       +/-   ##
===========================================
- Coverage   44.74%   14.47%   -30.27%     
===========================================
  Files         621      879      +258     
  Lines       55994    82527    +26533     
===========================================
- Hits        25054    11948    -13106     
- Misses      30940    70579    +39639     
Flag Coverage Δ
test_integration_espnet2 ?
test_integration_espnetez ?
test_python_espnetez 12.72% <ø> (?)
test_utils 20.63% <ø> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@Jungjee
Copy link
Contributor

Jungjee commented Jun 6, 2025

LGTM.
Adding comments to convert_trial.py, generate_trial.py, data_prep.py would be recommended, if possible.

@sw005320 sw005320 merged commit 2d1574b into espnet:master Jun 12, 2025
38 checks passed
@sw005320
Copy link
Contributor

Thanks @holvan!
I just merged this PR, but please create the follow-up PR for the tuned models.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants