update model card README.md
Browse files
README.md
CHANGED
|
@@ -1,10 +1,6 @@
|
|
| 1 |
---
|
| 2 |
-
language:
|
| 3 |
-
- tr
|
| 4 |
license: apache-2.0
|
| 5 |
tags:
|
| 6 |
-
- automatic-speech-recognition
|
| 7 |
-
- common_voice
|
| 8 |
- generated_from_trainer
|
| 9 |
datasets:
|
| 10 |
- common_voice
|
|
@@ -18,11 +14,11 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 18 |
|
| 19 |
#
|
| 20 |
|
| 21 |
-
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the
|
| 22 |
It achieves the following results on the evaluation set:
|
| 23 |
-
- Loss: 0.
|
| 24 |
-
- Wer: 0.
|
| 25 |
-
- Cer: 0.
|
| 26 |
|
| 27 |
## Model description
|
| 28 |
|
|
@@ -48,11 +44,34 @@ The following hyperparameters were used during training:
|
|
| 48 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 49 |
- lr_scheduler_type: linear
|
| 50 |
- lr_scheduler_warmup_steps: 500
|
| 51 |
-
- num_epochs:
|
| 52 |
- mixed_precision_training: Native AMP
|
| 53 |
|
| 54 |
### Training results
|
| 55 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
|
| 57 |
|
| 58 |
### Framework versions
|
|
|
|
| 1 |
---
|
|
|
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
tags:
|
|
|
|
|
|
|
| 4 |
- generated_from_trainer
|
| 5 |
datasets:
|
| 6 |
- common_voice
|
|
|
|
| 14 |
|
| 15 |
#
|
| 16 |
|
| 17 |
+
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the common_voice dataset.
|
| 18 |
It achieves the following results on the evaluation set:
|
| 19 |
+
- Loss: 0.7360
|
| 20 |
+
- Wer: 0.4754
|
| 21 |
+
- Cer: 0.1339
|
| 22 |
|
| 23 |
## Model description
|
| 24 |
|
|
|
|
| 44 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 45 |
- lr_scheduler_type: linear
|
| 46 |
- lr_scheduler_warmup_steps: 500
|
| 47 |
+
- num_epochs: 100.0
|
| 48 |
- mixed_precision_training: Native AMP
|
| 49 |
|
| 50 |
### Training results
|
| 51 |
|
| 52 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|
| 53 |
+
|:-------------:|:-----:|:-----:|:---------------:|:------:|:------:|
|
| 54 |
+
| 1.0779 | 4.59 | 500 | 0.8260 | 0.7395 | 0.2354 |
|
| 55 |
+
| 0.7573 | 9.17 | 1000 | 0.7544 | 0.6960 | 0.2100 |
|
| 56 |
+
| 0.8225 | 13.76 | 1500 | 0.6867 | 0.6672 | 0.2021 |
|
| 57 |
+
| 0.621 | 18.35 | 2000 | 0.6824 | 0.6209 | 0.1874 |
|
| 58 |
+
| 0.6362 | 22.94 | 2500 | 0.6712 | 0.6286 | 0.1904 |
|
| 59 |
+
| 0.624 | 27.52 | 3000 | 0.6940 | 0.6116 | 0.1820 |
|
| 60 |
+
| 0.4781 | 32.11 | 3500 | 0.6966 | 0.5989 | 0.1735 |
|
| 61 |
+
| 0.5685 | 36.7 | 4000 | 0.6742 | 0.5971 | 0.1769 |
|
| 62 |
+
| 0.4384 | 41.28 | 4500 | 0.6904 | 0.5999 | 0.1767 |
|
| 63 |
+
| 0.5509 | 45.87 | 5000 | 0.6734 | 0.5641 | 0.1692 |
|
| 64 |
+
| 0.3665 | 50.46 | 5500 | 0.7018 | 0.5662 | 0.1680 |
|
| 65 |
+
| 0.3914 | 55.05 | 6000 | 0.7121 | 0.5552 | 0.1631 |
|
| 66 |
+
| 0.2467 | 59.63 | 6500 | 0.6657 | 0.5374 | 0.1563 |
|
| 67 |
+
| 0.2576 | 64.22 | 7000 | 0.6920 | 0.5316 | 0.1554 |
|
| 68 |
+
| 0.2711 | 68.81 | 7500 | 0.6900 | 0.5176 | 0.1495 |
|
| 69 |
+
| 0.2626 | 73.39 | 8000 | 0.6843 | 0.5043 | 0.1454 |
|
| 70 |
+
| 0.1377 | 77.98 | 8500 | 0.7383 | 0.5101 | 0.1470 |
|
| 71 |
+
| 0.2005 | 82.57 | 9000 | 0.7228 | 0.5045 | 0.1430 |
|
| 72 |
+
| 0.1355 | 87.16 | 9500 | 0.7231 | 0.4869 | 0.1375 |
|
| 73 |
+
| 0.0431 | 91.74 | 10000 | 0.7397 | 0.4749 | 0.1350 |
|
| 74 |
+
| 0.0586 | 96.33 | 10500 | 0.7360 | 0.4754 | 0.1339 |
|
| 75 |
|
| 76 |
|
| 77 |
### Framework versions
|