Model card: link STT-API and Semantic-VAD repos
Browse files
README.md
CHANGED
|
@@ -133,6 +133,10 @@ def p_end_of_turn(pcm: np.ndarray) -> float:
|
|
| 133 |
return float(sess.run(None, {"input_features": feats})[0].reshape(-1)[0])
|
| 134 |
```
|
| 135 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 136 |
Use `p ≥ 0.3` as "the turn is over" for the operating point in the table above. The PyTorch loading
|
| 137 |
snippet (a `WhisperEncoder` subclass that accepts the 8 s window + the 3-layer head) is on the tiny model's
|
| 138 |
card and works unchanged with this repo id (`d_model` 512).
|
|
|
|
| 133 |
return float(sess.run(None, {"input_features": feats})[0].reshape(-1)[0])
|
| 134 |
```
|
| 135 |
|
| 136 |
+
For LiveKit Agents use it as the backend of [STT-API](https://github.com/Scicom-AI-Enterprise-Organization/STT-API)'s
|
| 137 |
+
`SemanticVAD` through `WhisperEoTOnnx` from the
|
| 138 |
+
[Semantic-VAD](https://github.com/Scicom-AI-Enterprise-Organization/Semantic-VAD) repo, exactly as on the tiny model's card.
|
| 139 |
+
|
| 140 |
Use `p ≥ 0.3` as "the turn is over" for the operating point in the table above. The PyTorch loading
|
| 141 |
snippet (a `WhisperEncoder` subclass that accepts the 8 s window + the 3-layer head) is on the tiny model's
|
| 142 |
card and works unchanged with this repo id (`d_model` 512).
|