breco/whisper-small-few-audios

Whisper Small Few Audios - vfranchis

This model is a fine-tuned version of openai/whisper-small on the Few audios 1.0 dataset. It achieves the following results on the evaluation set:

• Loss: 1.6364
• Wer: 66.6667

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 10
• training_steps: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.6824	2.8571	10	1.6364	66.6667
4.2687	5.7143	20	1.6364	66.6667
2.6441	8.5714	30	1.6364	66.6667
1.8789	11.4286	40	1.6364	66.6667
1.3406	14.2857	50	1.6364	66.6667
0.8864	17.1429	60	1.6364	66.6667
1.0665	20.0	70	1.6364	66.6667
0.5324	22.8571	80	1.6364	66.6667
4.0741	25.7143	90	1.6364	66.6667
2.8755	28.5714	100	1.6364	66.6667

Framework versions

• Transformers 4.44.2
• Pytorch 2.3.1+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1