End of training

README.md

@@ -20,8 +20,8 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the AtCO2_UWB dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.6647
-- Wer: 34.2671
+- Loss: 0.5665
+- Wer: 26.4179
 
 ## Model description
 
@@ -41,21 +41,24 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 1e-05
-- train_batch_size: 8
-- eval_batch_size: 8
+- train_batch_size: 16
+- 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: 500
-- training_steps: 2000
+- num_epochs: 5
 - mixed_precision_training: Native AMP
 
 ### Training results
 
-| Training Loss | Epoch  | Step | Validation Loss | Wer     |
-|:-------------:|:------:|:----:|:---------------:|:-------:|
-| 0.8235        | 0.8217 | 1000 | 0.7798          | 44.8930 |
-| 0.5161        | 1.6434 | 2000 | 0.6647          | 34.2671 |
+| Training Loss | Epoch | Step | Validation Loss | Wer     |
+|:-------------:|:-----:|:----:|:---------------:|:-------:|
+| 0.7865        | 1.0   | 609  | 0.8728          | 52.9827 |
+| 0.5848        | 2.0   | 1218 | 0.6320          | 34.8257 |
+| 0.4175        | 3.0   | 1827 | 0.5784          | 29.2477 |
+| 0.2928        | 4.0   | 2436 | 0.5656          | 27.7268 |
+| 0.241         | 5.0   | 3045 | 0.5665          | 26.4179 |
 
 
 ### Framework versions