Model save

README.md

@@ -2,8 +2,6 @@
 license: apache-2.0
 base_model: facebook/wav2vec2-large-xlsr-53
 tags:
-- automatic-speech-recognition
-- DewiBrynJones/banc-trawsgrifiadau-bangor-clean-with-ccv
 - generated_from_trainer
 metrics:
 - wer
@@ -17,10 +15,10 @@ should probably proofread and complete it, then remove this comment. -->
 
 # wav2vec2-xlsr-53-ft-btb-ccv-cy
 
-This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the DEWIBRYNJONES/BANC-TRAWSGRIFIADAU-BANGOR-CLEAN-WITH-CCV - DEFAULT dataset.
+This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.5418
-- Wer: 0.4178
+- Loss: 0.7622
+- Wer: 0.5187
 
 ## Model description
 
@@ -40,34 +38,39 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 0.0003
-- train_batch_size: 16
+- train_batch_size: 8
 - eval_batch_size: 16
 - seed: 42
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
-- lr_scheduler_warmup_steps: 300
-- training_steps: 3000
+- lr_scheduler_warmup_steps: 500
+- training_steps: 10000
 - mixed_precision_training: Native AMP
 
 ### Training results
 
-| Training Loss | Epoch  | Step | Validation Loss | Wer    |
-|:-------------:|:------:|:----:|:---------------:|:------:|
-| No log        | 0.0672 | 200  | 3.1445          | 1.0    |
-| No log        | 0.1344 | 400  | 2.7407          | 1.0000 |
-| 4.0188        | 0.2016 | 600  | 1.2700          | 0.8484 |
-| 4.0188        | 0.2688 | 800  | 0.9953          | 0.7435 |
-| 1.0707        | 0.3360 | 1000 | 0.8647          | 0.6541 |
-| 1.0707        | 0.4032 | 1200 | 0.7889          | 0.5784 |
-| 1.0707        | 0.4704 | 1400 | 0.7465          | 0.5440 |
-| 0.8175        | 0.5376 | 1600 | 0.6828          | 0.5042 |
-| 0.8175        | 0.6048 | 1800 | 0.6549          | 0.4952 |
-| 0.7148        | 0.6720 | 2000 | 0.6290          | 0.4906 |
-| 0.7148        | 0.7392 | 2200 | 0.6113          | 0.4576 |
-| 0.7148        | 0.8065 | 2400 | 0.5719          | 0.4405 |
-| 0.6374        | 0.8737 | 2600 | 0.5644          | 0.4314 |
-| 0.6374        | 0.9409 | 2800 | 0.5483          | 0.4190 |
-| 0.6013        | 1.0081 | 3000 | 0.5418          | 0.4178 |
+| Training Loss | Epoch  | Step  | Validation Loss | Wer    |
+|:-------------:|:------:|:-----:|:---------------:|:------:|
+| 4.5088        | 0.0840 | 500   | 2.0263          | 1.0069 |
+| 1.2959        | 0.1680 | 1000  | 1.0948          | 0.7472 |
+| 0.9954        | 0.2521 | 1500  | 0.9249          | 0.6714 |
+| 0.8907        | 0.3361 | 2000  | 0.8197          | 0.6131 |
+| 0.8197        | 0.4201 | 2500  | 0.7610          | 0.5501 |
+| 0.7873        | 0.5041 | 3000  | 0.7496          | 0.5437 |
+| 0.748         | 0.5881 | 3500  | 0.6833          | 0.5240 |
+| 0.7133        | 0.6722 | 4000  | 0.6442          | 0.4790 |
+| 0.6809        | 0.7562 | 4500  | 0.6224          | 0.4703 |
+| 0.6517        | 0.8402 | 5000  | 0.5992          | 0.4571 |
+| 0.6363        | 0.9242 | 5500  | 0.5727          | 0.4373 |
+| 0.6111        | 1.0082 | 6000  | 0.5573          | 0.4200 |
+| 0.5352        | 1.0923 | 6500  | 0.5567          | 0.4259 |
+| 0.6231        | 1.1763 | 7000  | 0.8251          | 0.6300 |
+| 0.7995        | 1.2603 | 7500  | 0.6610          | 0.4686 |
+| 0.7696        | 1.3443 | 8000  | 0.7331          | 0.4870 |
+| 0.7978        | 1.4283 | 8500  | 0.7790          | 0.5432 |
+| 0.7844        | 1.5124 | 9000  | 0.8260          | 0.5482 |
+| 0.8211        | 1.5964 | 9500  | 0.8065          | 0.5491 |
+| 0.7744        | 1.6804 | 10000 | 0.7622          | 0.5187 |
 
 
 ### Framework versions