Downloads · 30 days
2
1% of all-time downloads
CLEAR-Global/w2v-bert-2.0-yoruba_naijavoices_1h
w2v-bert-2.0-yoruba_naijavoices_1h is a automatic speech recognition model from CLEAR-Global. Use it when you need speech turned into text. It is set up for transformers. The card lists the license as mit.
should probably proofread and complete it, then remove this comment. --
Downloads · 30 days
2
1% of all-time downloads
All-time downloads
148
Public
Parameters
606M
96.9 GB on disk
Likes
0
Public
Click a slice to open those files.
.pt4.8 GB · 50%
From the Hugging Face model README
This model is a fine-tuned version of facebook/w2v-bert-2.0 on the NAIJAVOICES_YORUBA_1H - NA dataset. It achieves the following results on the evaluation set:
More information needed
More information needed
More information needed
The following hyperparameters were used during training:
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 2.9463 | 25.0 | 100 | 5.9998 | 1.5303 | 0.9291 |
| 1.8808 | 50.0 | 200 | 3.7688 | 1.0 | 1.0 |
| 1.6931 | 75.0 | 300 | 3.3619 | 0.9969 | 0.9334 |
| 1.5762 | 100.0 | 400 | 3.1325 | 0.9995 | 0.9811 |
| 1.0234 | 125.0 | 500 | 1.9826 | 1.0017 | 0.6494 |
| 0.6432 | 150.0 | 600 | 1.3328 | 0.9127 | 0.4606 |
| 0.5951 | 175.0 | 700 | 1.1715 | 0.8537 | 0.4216 |
| 0.4745 | 200.0 | 800 | 1.0785 | 0.7823 | 0.3882 |
| 0.4669 | 225.0 | 900 | 1.0518 | 0.7507 | 0.3777 |
| 0.389 | 250.0 | 1000 | 1.0562 | 0.7474 | 0.3745 |
| 0.3123 | 275.0 | 1100 | 1.0829 | 0.7382 | 0.3673 |
| 0.2463 | 300.0 | 1200 | 1.1173 | 0.7319 | 0.3696 |
| 0.2244 | 325.0 | 1300 | 1.2219 | 0.7268 | 0.3678 |
| 0.1695 | 350.0 | 1400 | 1.2698 | 0.7419 | 0.3738 |