Whisper Zulu ASR Models
Collection
This is a collection of Whisper models for transcribing audio/video in the Zulu language.
•
4 items
•
Updated
•
1
whisper-zulu-medium
This model is a fine-tuned version of openai/whisper-medium on the wjbmattingly/zulu_merged_audio dataset. It achieves the following results on the evaluation set:
- Loss: 0.2949
- Wer: 0.1993
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- total_train_batch_size: 64
- total_eval_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 400
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.8378 | 1.25 | 100 | 0.7290 | 0.5605 |
| 0.3624 | 2.5 | 200 | 0.4048 | 0.2791 |
| 0.2279 | 3.75 | 300 | 0.3236 | 0.2187 |
| 0.1524 | 5.0 | 400 | 0.2949 | 0.1993 |
Framework versions
- Transformers 4.45.0.dev0
- Pytorch 2.0.1+cu117
- Datasets 2.21.0
- Tokenizers 0.19.1
- Downloads last month
- 23
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for TheirStory/whisper-medium-zulu
Base model
openai/whisper-mediumCollection including TheirStory/whisper-medium-zulu
Evaluation results
- Wer on wjbmattingly/zulu_merged_audioself-reported0.199