End of training

README.md

@@ -8,8 +8,6 @@ tags:
 datasets:
 - voice-biomarkers/openslr-32-hq-SA-languages-Afrikaans
 - google/fleurs
-- dsfsi-anv/multilingual-nchlt-dataset
-- andreoosthuizen/afrikaans-30s
 metrics:
 - wer
 model-index:
@@ -27,8 +25,9 @@ model-index:
     metrics:
     - name: Wer
       type: wer
-      value: 49.05627705627706
+      value: 52.17316017316017
 ---
+
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 should probably proofread and complete it, then remove this comment. -->
 
@@ -36,9 +35,9 @@ 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 Common Voice 17.0 dataset.
 It achieves the following results on the evaluation set:
-- Loss: 1.3265
-- Wer: 49.0563
-- Cer: 20.0715
+- Loss: 1.3668
+- Wer: 52.1732
+- Cer: 20.9395
 
 ## Model description
 
@@ -58,23 +57,38 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 1e-05
-- train_batch_size: 128
-- eval_batch_size: 128
+- train_batch_size: 64
+- eval_batch_size: 64
 - seed: 42
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_ratio: 0.04
-- training_steps: 5000
+- training_steps: 2000
 
 ### Training results
 
-| Training Loss | Epoch | Step | Validation Loss | Wer     | Cer     |
-|:-------------:|:-----:|:----:|:---------------:|:-------:|:-------:|
-| 0.0873        | 0.2   | 1000 | 1.1737          | 43.8442 | 18.5673 |
-| 0.0323        | 1.014 | 2000 | 1.2562          | 46.8571 | 19.0335 |
-| 0.0184        | 1.214 | 3000 | 1.2972          | 42.9957 | 17.3357 |
-| 0.0164        | 2.028 | 4000 | 1.3082          | 45.8528 | 18.4676 |
-| 0.0154        | 2.228 | 5000 | 1.3265          | 49.0563 | 20.0715 |
+| Training Loss | Epoch   | Step | Validation Loss | Wer     | Cer     |
+|:-------------:|:-------:|:----:|:---------------:|:-------:|:-------:|
+| 1.3199        | 0.05    | 100  | 1.4770          | 59.2208 | 24.5022 |
+| 0.6393        | 1.0315  | 200  | 1.2510          | 51.2554 | 20.9454 |
+| 0.3811        | 2.013   | 300  | 1.2197          | 49.4545 | 20.1155 |
+| 0.261         | 2.063   | 400  | 1.2089          | 48.3290 | 19.2036 |
+| 0.186         | 3.0445  | 500  | 1.2141          | 48.0693 | 19.8575 |
+| 0.1459        | 4.026   | 600  | 1.2341          | 49.8701 | 20.2621 |
+| 0.0963        | 5.0075  | 700  | 1.2517          | 48.4675 | 19.5437 |
+| 0.0809        | 5.0575  | 800  | 1.2674          | 51.0823 | 21.0715 |
+| 0.0536        | 6.039   | 900  | 1.2812          | 48.2597 | 19.5408 |
+| 0.0432        | 7.0205  | 1000 | 1.3003          | 48.5022 | 19.4910 |
+| 0.0379        | 8.002   | 1100 | 1.3117          | 51.6190 | 21.2298 |
+| 0.0333        | 8.052   | 1200 | 1.3314          | 52.3463 | 21.7078 |
+| 0.0247        | 9.0335  | 1300 | 1.3389          | 52.0    | 21.4644 |
+| 0.0201        | 10.015  | 1400 | 1.3484          | 51.4113 | 22.1769 |
+| 0.0194        | 10.065  | 1500 | 1.3469          | 51.8442 | 21.0685 |
+| 0.0191        | 11.0465 | 1600 | 1.3536          | 52.4502 | 21.3471 |
+| 0.0179        | 12.028  | 1700 | 1.3611          | 51.7229 | 21.1272 |
+| 0.0155        | 13.0095 | 1800 | 1.3637          | 52.4329 | 20.9512 |
+| 0.0159        | 13.0595 | 1900 | 1.3651          | 52.0346 | 20.9483 |
+| 0.0152        | 14.041  | 2000 | 1.3668          | 52.1732 | 20.9395 |
 
 
 ### Framework versions
@@ -83,16 +97,3 @@ The following hyperparameters were used during training:
 - Pytorch 2.3.0+cu121
 - Datasets 2.19.1
 - Tokenizers 0.19.1
-
-## Citation
-
-Please cite the model using the following BibTeX entry:
-
-```bibtex
-@misc{deepdml/whisper-tiny-af-mix-norm,
-      title={Fine-tuned Whisper tiny ASR model for speech recognition in Afrikaans},
-      author={Jimenez, David},
-      howpublished={\url{https://huggingface.co/deepdml/whisper-tiny-af-mix-norm}},
-      year={2026}
-    }
-```