Commit
•
8461387
1
Parent(s):
c6f8a39
Update README.md
Browse files
README.md
CHANGED
|
@@ -7,4 +7,32 @@ language:
|
|
| 7 |
metrics:
|
| 8 |
- wer
|
| 9 |
library_name: transformers
|
| 10 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
metrics:
|
| 8 |
- wer
|
| 9 |
library_name: transformers
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
The model is a fine-tuned version of Meta model [facebook/wav2vec2-large-lv60](https://huggingface.co/facebook/wav2vec2-large-lv60) on timit dataset. It yields aLos result the following:
|
| 13 |
+
WER: 0.247330
|
| 14 |
+
Validation Loss: 0.299390
|
| 15 |
+
> The intention behind is to make the model Multilingual although yet it is just for english
|
| 16 |
+
### Training result
|
| 17 |
+
| Step | Training Loss | Validation Loss | Wer |
|
| 18 |
+
|------|---------------|-----------------|----------|
|
| 19 |
+
| 500 | 4.808300 | 2.929884 | 1.000000 |
|
| 20 |
+
| 1000 | 1.543200 | 0.367235 | 0.415202 |
|
| 21 |
+
| 1500 | 0.305000 | 0.264133 | 0.321274 |
|
| 22 |
+
| 2000 | 0.161400 | 0.268306 | 0.294949 |
|
| 23 |
+
| 2500 | 0.109600 | 0.258425 | 0.276135 |
|
| 24 |
+
| 3000 | 0.084900 | 0.263341 | 0.268348 |
|
| 25 |
+
| 3500 | 0.066600 | 0.258404 | 0.259734 |
|
| 26 |
+
| 4000 | 0.055800 | 0.257602 | 0.249397 |
|
| 27 |
+
| 4500 | 0.045500 | 0.271701 | 0.253118 |
|
| 28 |
+
| 5000 | 0.042200 | 0.292905 | 0.251327 |
|
| 29 |
+
| 5500 | 0.038300 | 0.278083 | 0.248019 |
|
| 30 |
+
| 6000 | 0.033000 | 0.295660 | 0.250362 |
|
| 31 |
+
| 6500 | 0.031400 | 0.286542 | 0.248639 |
|
| 32 |
+
| 7000 | 0.030800 | 0.299390 | 0.247330 |
|
| 33 |
+
|
| 34 |
+
### Training Procedure
|
| 35 |
+
#### Hyper parameters
|
| 36 |
+
More information to come
|
| 37 |
+
### Tensorboard
|
| 38 |
+
More information to come
|