Add license, pipeline tag, and library name metadata
Browse filesThis PR updates the model card by adding essential metadata:
- The `license: apache-2.0` for proper attribution and legal clarity, verified via the linked PyPI project page.
- The `pipeline_tag: audio-to-audio` to correctly categorize the model and improve discoverability on the Hugging Face Hub.
- The `library_name: dualcodec`, as evidenced by the programmatic usage shown in the README (e.g., `import dualcodec`), enabling automated code snippets on the Hub.
README.md
CHANGED
|
@@ -1,4 +1,8 @@
|
|
| 1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
# DualCodec: A Low-Frame-Rate, Semantically-Enhanced Neural Audio Codec for Speech Generation
|
| 4 |
|
|
@@ -194,7 +198,10 @@ data.segment_speech.segment_length=24000
|
|
| 194 |
|
| 195 |
```
|
| 196 |
|
|
|
|
| 197 |
|
|
|
|
|
|
|
| 198 |
|
| 199 |
## Finetuning DualCodec
|
| 200 |
1. Install other necessary components for training:
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
pipeline_tag: audio-to-audio
|
| 4 |
+
library_name: dualcodec
|
| 5 |
+
---
|
| 6 |
|
| 7 |
# DualCodec: A Low-Frame-Rate, Semantically-Enhanced Neural Audio Codec for Speech Generation
|
| 8 |
|
|
|
|
| 198 |
|
| 199 |
```
|
| 200 |
|
| 201 |
+
Below are example training loss curves for a DualCodec 25hz model:
|
| 202 |
|
| 203 |
+
|
| 204 |
+
|
| 205 |
|
| 206 |
## Finetuning DualCodec
|
| 207 |
1. Install other necessary components for training:
|