metadata

library_name: transformers
language:
  - hi
license: apache-2.0
base_model: openai/whisper-medium
tags:
  - hf-asr-leaderboard
  - generated_from_trainer
datasets:
  - mozilla-foundation/common_voice_11_0
metrics:
  - wer
model-index:
  - name: Whisper Swahili Medium
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Common Voice 11.0
          type: mozilla-foundation/common_voice_11_0
          config: sw
          split: None
          args: 'config: hi, split: test'
        metrics:
          - name: Wer
            type: wer
            value: 25.261512288203786

Whisper Swahili Medium

This model is a fine-tuned version of openai/whisper-medium on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

Loss: 0.3519
Wer: 25.2615

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: Use OptimizerNames.ADAMW_TORCH 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: 500
training_steps: 4000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3595	0.4342	1000	0.4586	31.5976
0.3001	0.8684	2000	0.3794	27.8295
0.1451	1.3026	3000	0.3701	26.1972
0.1469	1.7369	4000	0.3519	25.2615

Framework versions

Transformers 4.46.0.dev0
Pytorch 2.1.1+cu121
Datasets 3.0.1
Tokenizers 0.20.0