bkh6722
/

bach-arb

@@ -13,8 +13,8 @@ should probably proofread and complete it, then remove this comment. -->
 This model is a fine-tuned version of [jonatasgrosman/wav2vec2-large-xlsr-53-german](https://huggingface.co/jonatasgrosman/wav2vec2-large-xlsr-53-german) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 1.1195
-- Wer: 0.1773
 ## Model description
@@ -42,29 +42,29 @@ The following hyperparameters were used during training:
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_steps: 500
-- num_epochs: 70
 - mixed_precision_training: Native AMP
 ### Training results
 | Training Loss | Epoch | Step | Validation Loss | Wer    |
 |:-------------:|:-----:|:----:|:---------------:|:------:|
-| 21.636        | 4.34  | 100  | 3.0908          | 1.0    |
-| 2.7682        | 8.68  | 200  | 1.9050          | 0.9553 |
-| 1.2501        | 13.04 | 300  | 0.7659          | 0.3830 |
-| 0.4483        | 17.38 | 400  | 0.9268          | 0.2683 |
-| 0.2845        | 21.72 | 500  | 0.7930          | 0.2072 |
-| 0.2031        | 26.09 | 600  | 0.9293          | 0.1997 |
-| 0.1386        | 30.43 | 700  | 0.9864          | 0.1923 |
-| 0.1047        | 34.77 | 800  | 1.0561          | 0.1952 |
-| 0.0698        | 39.13 | 900  | 1.0644          | 0.1893 |
-| 0.0547        | 43.47 | 1000 | 0.9583          | 0.1893 |
-| 0.0428        | 47.81 | 1100 | 1.0994          | 0.1848 |
-| 0.0296        | 52.17 | 1200 | 1.1003          | 0.1893 |
-| 0.0253        | 56.51 | 1300 | 1.1177          | 0.1923 |
-| 0.0216        | 60.85 | 1400 | 1.1159          | 0.1610 |
-| 0.0136        | 65.21 | 1500 | 1.1248          | 0.1759 |
-| 0.0133        | 69.55 | 1600 | 1.1195          | 0.1773 |
 ### Framework versions

 This model is a fine-tuned version of [jonatasgrosman/wav2vec2-large-xlsr-53-german](https://huggingface.co/jonatasgrosman/wav2vec2-large-xlsr-53-german) on an unknown dataset.
 It achieves the following results on the evaluation set:
+- Loss: 8.0485
+- Wer: 1.1740
 ## Model description
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_steps: 500
+- num_epochs: 320
 - mixed_precision_training: Native AMP
 ### Training results
 | Training Loss | Epoch | Step | Validation Loss | Wer    |
 |:-------------:|:-----:|:----:|:---------------:|:------:|
+| 6.61          | 20.0  | 100  | 3.0666          | 1.0    |
+| 2.6693        | 40.0  | 200  | 3.0447          | 0.9971 |
+| 1.7005        | 60.0  | 300  | 4.0288          | 1.0118 |
+| 0.7894        | 80.0  | 400  | 5.1604          | 1.0265 |
+| 0.3588        | 100.0 | 500  | 6.2627          | 1.0560 |
+| 0.1847        | 120.0 | 600  | 6.1311          | 1.0708 |
+| 0.1079        | 140.0 | 700  | 6.5706          | 1.0914 |
+| 0.0723        | 160.0 | 800  | 7.3875          | 1.1976 |
+| 0.0506        | 180.0 | 900  | 7.0595          | 1.1740 |
+| 0.0378        | 200.0 | 1000 | 7.5306          | 1.1445 |
+| 0.041         | 220.0 | 1100 | 7.4536          | 1.0767 |
+| 0.0254        | 240.0 | 1200 | 7.7803          | 1.1504 |
+| 0.0211        | 260.0 | 1300 | 7.9156          | 1.1003 |
+| 0.0175        | 280.0 | 1400 | 8.0091          | 1.1681 |
+| 0.0141        | 300.0 | 1500 | 8.0537          | 1.1475 |
+| 0.0132        | 320.0 | 1600 | 8.0485          | 1.1740 |
 ### Framework versions