End of training

README.md

@@ -16,7 +16,7 @@ model-index:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
     dataset:
-      name: Fleurs
+      name: Common Voice 17.0
       type: google/fleurs
       config: af_za
       split: test
@@ -24,8 +24,9 @@ model-index:
     metrics:
     - name: Wer
       type: wer
-      value: 49.76623376623376
+      value: 49.05627705627706
 ---
+
 <!-- 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. -->
 
@@ -33,9 +34,9 @@ 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: 1.3003
-- Wer: 49.7662
-- Cer: 21.6198
+- Loss: 1.2724
+- Wer: 49.0563
+- Cer: 20.8105
 
 ## Model description
 
@@ -61,22 +62,20 @@ 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_ratio: 0.04
-- training_steps: 1000
+- training_steps: 800
 
 ### Training results
 
-| Training Loss | Epoch  | Step | Validation Loss | Wer     | Cer     |
-|:-------------:|:------:|:----:|:---------------:|:-------:|:-------:|
-| 1.2064        | 1.035  | 100  | 1.4604          | 58.8918 | 24.7544 |
-| 0.6353        | 3.005  | 200  | 1.2920          | 52.7273 | 21.1242 |
-| 0.4014        | 4.04   | 300  | 1.2630          | 50.6320 | 20.0833 |
-| 0.2765        | 6.01   | 400  | 1.2550          | 51.0649 | 20.7548 |
-| 0.2025        | 7.045  | 500  | 1.2630          | 49.4892 | 19.7754 |
-| 0.1471        | 9.015  | 600  | 1.2717          | 49.1948 | 20.8809 |
-| 0.1213        | 10.05  | 700  | 1.2860          | 49.2814 | 20.8017 |
-| 0.0935        | 12.02  | 800  | 1.2944          | 49.1602 | 20.8838 |
-| 0.0879        | 13.055 | 900  | 1.2994          | 51.5152 | 20.8457 |
-| 0.079         | 15.025 | 1000 | 1.3003          | 49.7662 | 21.6198 |
+| Training Loss | Epoch   | Step | Validation Loss | Wer     | Cer     |
+|:-------------:|:-------:|:----:|:---------------:|:-------:|:-------:|
+| 1.1777        | 1.0437  | 100  | 1.4515          | 58.3377 | 23.8894 |
+| 0.6382        | 3.0063  | 200  | 1.2939          | 52.7100 | 20.6170 |
+| 0.4144        | 4.05    | 300  | 1.2647          | 50.8571 | 20.1595 |
+| 0.297         | 6.0125  | 400  | 1.2563          | 52.0693 | 20.8838 |
+| 0.229         | 7.0563  | 500  | 1.2615          | 51.3939 | 20.8164 |
+| 0.1789        | 9.0188  | 600  | 1.2661          | 49.5584 | 20.9688 |
+| 0.1623        | 10.0625 | 700  | 1.2730          | 49.2468 | 20.9600 |
+| 0.1438        | 12.025  | 800  | 1.2724          | 49.0563 | 20.8105 |
 
 
 ### Framework versions
@@ -85,16 +84,3 @@ 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-fleurs-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-fleurs-norm}},
-      year={2026}
-    }
-```