Update README.md
Browse files
README.md
CHANGED
|
@@ -29,27 +29,14 @@ model-index:
|
|
| 29 |
- name: Test CER
|
| 30 |
type: cer
|
| 31 |
value: 2.61
|
| 32 |
-
|
| 33 |
-
name: Automatic Speech Recognition
|
| 34 |
-
type: automatic-speech-recognition
|
| 35 |
-
dataset:
|
| 36 |
-
name: Robust Speech Event - Dev Data
|
| 37 |
-
type: speech-recognition-community-v2/dev_data
|
| 38 |
-
args: fi
|
| 39 |
-
metrics:
|
| 40 |
-
- name: Test WER
|
| 41 |
-
type: wer
|
| 42 |
-
value: 11.533
|
| 43 |
-
- name: Test CER
|
| 44 |
-
type: cer
|
| 45 |
-
value: 2.611
|
| 46 |
---
|
| 47 |
# sammy786/wav2vec2-xlsr-czech
|
| 48 |
|
| 49 |
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - cs dataset.
|
| 50 |
It achieves the following results on evaluation set (which is 10 percent of train data set merged with other and dev datasets):
|
| 51 |
-
- Loss:
|
| 52 |
-
- Wer:
|
| 53 |
|
| 54 |
## Model description
|
| 55 |
"facebook/wav2vec2-xls-r-1b" was finetuned.
|
|
@@ -80,6 +67,40 @@ The following hyperparameters were used during training:
|
|
| 80 |
- mixed_precision_training: Native AMP
|
| 81 |
|
| 82 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 83 |
|
| 84 |
### Framework versions
|
| 85 |
- Transformers 4.16.0.dev0
|
|
|
|
| 29 |
- name: Test CER
|
| 30 |
type: cer
|
| 31 |
value: 2.61
|
| 32 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
---
|
| 34 |
# sammy786/wav2vec2-xlsr-czech
|
| 35 |
|
| 36 |
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - cs dataset.
|
| 37 |
It achieves the following results on evaluation set (which is 10 percent of train data set merged with other and dev datasets):
|
| 38 |
+
- Loss: 7.26
|
| 39 |
+
- Wer: 19.32
|
| 40 |
|
| 41 |
## Model description
|
| 42 |
"facebook/wav2vec2-xls-r-1b" was finetuned.
|
|
|
|
| 67 |
- mixed_precision_training: Native AMP
|
| 68 |
|
| 69 |
|
| 70 |
+
### Training results
|
| 71 |
+
|
| 72 |
+
| Step | Training Loss | Validation Loss | Wer |
|
| 73 |
+
|:----:|:-------------:|:---------------:|:--------:|
|
| 74 |
+
| 200 | 6.654600 | 3.329486 | 1.000000 |
|
| 75 |
+
| 400 | 1.700600 | 0.317266 | 0.409446 |
|
| 76 |
+
| 600 | 0.767400 | 0.211371 | 0.313981 |
|
| 77 |
+
| 800 | 0.718600 | 0.167771 | 0.280676 |
|
| 78 |
+
| 1000 | 0.661700 | 0.142229 | 0.258938 |
|
| 79 |
+
| 1200 | 0.594400 | 0.137321 | 0.256275 |
|
| 80 |
+
| 1400 | 0.583900 | 0.132922 | 0.248418 |
|
| 81 |
+
| 1600 | 0.565100 | 0.117214 | 0.238640 |
|
| 82 |
+
| 1800 | 0.369600 | 0.116954 | 0.238291 |
|
| 83 |
+
| 2000 | 0.292800 | 0.109973 | 0.227509 |
|
| 84 |
+
| 2200 | 0.255400 | 0.104955 | 0.228120 |
|
| 85 |
+
| 2400 | 0.266800 | 0.097268 | 0.220525 |
|
| 86 |
+
| 2600 | 0.232700 | 0.096055 | 0.213584 |
|
| 87 |
+
| 2800 | 0.213700 | 0.097770 | 0.218866 |
|
| 88 |
+
| 3000 | 0.209900 | 0.091633 | 0.210485 |
|
| 89 |
+
| 3200 | 0.196800 | 0.090342 | 0.208739 |
|
| 90 |
+
| 3400 | 0.200500 | 0.082326 | 0.204767 |
|
| 91 |
+
| 3600 | 0.176800 | 0.085491 | 0.204068 |
|
| 92 |
+
| 3800 | 0.170000 | 0.081289 | 0.201231 |
|
| 93 |
+
| 4000 | 0.166200 | 0.080762 | 0.200227 |
|
| 94 |
+
| 4200 | 0.161700 | 0.076671 | 0.198001 |
|
| 95 |
+
| 4400 | 0.147000 | 0.077383 | 0.196997 |
|
| 96 |
+
| 4600 | 0.141900 | 0.076057 | 0.195862 |
|
| 97 |
+
| 4800 | 0.144800 | 0.074612 | 0.195120 |
|
| 98 |
+
| 5000 | 0.138900 | 0.073138 | 0.193985 |
|
| 99 |
+
| 5200 | 0.143900 | 0.072802 | 0.192894 |
|
| 100 |
+
| 5400 | 0.131100 | 0.072764 | 0.193723 |
|
| 101 |
+
| 5600 | 0.137000 | 0.072697 | 0.193679 |
|
| 102 |
+
| 5800 | 0.133300 | 0.072651 | 0.193286 |
|
| 103 |
+
|
| 104 |
|
| 105 |
### Framework versions
|
| 106 |
- Transformers 4.16.0.dev0
|