whisper-medium-common_voice_17_0-id-10000

This model is a fine-tuned version of openai/whisper-medium on the mozilla-foundation/common_voice_17_0 id dataset. It achieves the following results on the evaluation set:

• Loss: 0.0574
• Wer: 0.0424

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.132	0.8457	1000	0.0963	0.0747
0.0503	1.6913	2000	0.0664	0.0526
0.023	2.5370	3000	0.0628	0.0727
0.011	3.3827	4000	0.0593	0.0437
0.0033	4.2283	5000	0.0575	0.0407
0.0017	5.0740	6000	0.0574	0.0448
0.0013	5.9197	7000	0.0554	0.0386
0.002	6.7653	8000	0.0555	0.0426
0.0002	7.6110	9000	0.0571	0.0421
0.0005	8.4567	10000	0.0574	0.0424

Framework versions

• Transformers 4.42.0.dev0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1