| language: | |
| - sw | |
| license: apache-2.0 | |
| tags: | |
| - whisper-asr | |
| - generated_from_trainer | |
| datasets: | |
| - mozilla-foundation/common_voice_11_0 | |
| model-index: | |
| - name: Whisper Small Swahili - Passie | |
| results: [] | |
| <!-- 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 Swahili - Passie | |
| This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the Common Voice 11.0 dataset. | |
| ## 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: 8 | |
| - seed: 42 | |
| - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 | |
| - lr_scheduler_type: linear | |
| - lr_scheduler_warmup_steps: 5 | |
| - training_steps: 40 | |
| - mixed_precision_training: Native AMP | |
| ### Training results | |
| ### Framework versions | |
| - Transformers 4.30.0.dev0 | |
| - Pytorch 2.0.0+cu118 | |
| - Datasets 2.12.0 | |
| - Tokenizers 0.13.3 | |