Update banner and fix README metadata
Browse files- README.md +20 -0
- banner.svg +86 -0
README.md
CHANGED
|
@@ -1,5 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
# ECAPA-QAT
|
| 2 |
|
|
|
|
|
|
|
| 3 |
**Quantization-Aware Trained ECAPA-TDNN for Speaker Verification**
|
| 4 |
|
| 5 |
A mixed-precision W(4/8)A32 speaker embedding model trained with a 5-phase progressive QAT strategy and cosine distillation. Achieves **2.61% EER** on VoxCeleb1-O while fitting in **4 MB** on disk and **7.6 MB** in RAM — making it suitable for CPU-only servers, edge devices, and mobile deployment.
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
tags:
|
| 6 |
+
- speaker-verification
|
| 7 |
+
- speaker-recognition
|
| 8 |
+
- ecapa-tdnn
|
| 9 |
+
- quantization
|
| 10 |
+
- qat
|
| 11 |
+
- mixed-precision
|
| 12 |
+
- edge
|
| 13 |
+
datasets:
|
| 14 |
+
- voxceleb2
|
| 15 |
+
metrics:
|
| 16 |
+
- eer
|
| 17 |
+
---
|
| 18 |
+
|
| 19 |
# ECAPA-QAT
|
| 20 |
|
| 21 |
+
|
| 22 |
+
|
| 23 |
**Quantization-Aware Trained ECAPA-TDNN for Speaker Verification**
|
| 24 |
|
| 25 |
A mixed-precision W(4/8)A32 speaker embedding model trained with a 5-phase progressive QAT strategy and cosine distillation. Achieves **2.61% EER** on VoxCeleb1-O while fitting in **4 MB** on disk and **7.6 MB** in RAM — making it suitable for CPU-only servers, edge devices, and mobile deployment.
|
banner.svg
ADDED
|