Add model card

README.md

@@ -1,77 +1,41 @@
 ---
-library_name: transformers
+language: zu
 license: apache-2.0
-base_model: openai/whisper-small
 tags:
-- generated_from_trainer
-metrics:
-- wer
-model-index:
-- name: whisper-small-isizulu-noisy
-  results: []
+- whisper
+- automatic-speech-recognition
+- south-african-languages
+datasets:
+- zionia/isizulu-asr-train
+- zionia/isizulu-asr-gaussian-noise
+base_model: openai/whisper-small
 ---
 
-<!-- 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. -->
-
-# whisper-small-isizulu-noisy
-
-This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the None dataset.
-It achieves the following results on the evaluation set:
-- Loss: 0.9578
-- Wer: 77.9164
-
-## Model description
-
-More information needed
-
-## Intended uses & limitations
-
-More information needed
-
-## Training and evaluation data
+# Whisper Small - Combined South African Languages
 
-More information needed
+This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on multiple South African language datasets.
 
-## Training procedure
+## Training Data
 
-### Training hyperparameters
+This model was trained on 2 combined datasets:
+- zionia/isizulu-asr-train
+- zionia/isizulu-asr-gaussian-noise
 
-The following hyperparameters were used during training:
-- learning_rate: 1e-05
-- train_batch_size: 32
-- eval_batch_size: 16
-- seed: 42
-- optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
-- lr_scheduler_type: linear
-- lr_scheduler_warmup_steps: 100
-- num_epochs: 15
-- mixed_precision_training: Native AMP
+Total training samples: 1050
+Total test samples: 263
 
-### Training results
+## Training Details
 
-| Training Loss | Epoch | Step | Validation Loss | Wer      |
-|:-------------:|:-----:|:----:|:---------------:|:--------:|
-| 2.7094        | 1.0   | 33   | 2.2550          | 108.7960 |
-| 1.5701        | 2.0   | 66   | 1.4053          | 90.1435  |
-| 1.0347        | 3.0   | 99   | 1.0760          | 103.7430 |
-| 0.5594        | 4.0   | 132  | 0.9268          | 100.6238 |
-| 0.2742        | 5.0   | 165  | 0.9051          | 80.2246  |
-| 0.1504        | 6.0   | 198  | 0.8892          | 89.5820  |
-| 0.065         | 7.0   | 231  | 0.9149          | 81.1603  |
-| 0.0369        | 8.0   | 264  | 0.9126          | 66.1260  |
-| 0.0215        | 9.0   | 297  | 0.9266          | 66.9370  |
-| 0.0129        | 10.0  | 330  | 0.9321          | 63.5683  |
-| 0.0076        | 11.0  | 363  | 0.9407          | 66.1884  |
-| 0.006         | 12.0  | 396  | 0.9498          | 64.3793  |
-| 0.0051        | 13.0  | 429  | 0.9536          | 77.7293  |
-| 0.0046        | 14.0  | 462  | 0.9570          | 77.9164  |
-| 0.0044        | 15.0  | 495  | 0.9578          | 77.9164  |
+- **Training epochs:** 15
+- **Learning rate:** 1e-05
+- **Batch size:** 16
+- **Best WER:** 63.57%
 
+## Usage
 
-### Framework versions
+```python
+from transformers import pipeline
 
-- Transformers 4.57.0
-- Pytorch 2.8.0+cu128
-- Datasets 4.2.0
-- Tokenizers 0.22.1
+pipe = pipeline("automatic-speech-recognition", model="zionia/whisper-small-isizulu-noisy")
+result = pipe("path/to/audio.wav")
+```