whisper-tiny-common_voice_17_0-id

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

• Loss: 0.2000
• Wer: 0.1807

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: 500
• training_steps: 20000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4911	0.4229	1000	0.4546	0.3321
0.4078	0.8458	2000	0.3520	0.2807
0.2679	1.2688	3000	0.3050	0.2421
0.2423	1.6917	4000	0.2725	0.2217
0.169	2.1146	5000	0.2515	0.2184
0.1646	2.5375	6000	0.2377	0.2082
0.1731	2.9605	7000	0.2189	0.1911
0.1017	3.3834	8000	0.2135	0.1970
0.0985	3.8063	9000	0.2077	0.1819
0.0828	4.2292	10000	0.2070	0.1792
0.06	4.6521	11000	0.1991	0.1826
0.0629	5.0751	12000	0.2012	0.1918
0.0545	5.4980	13000	0.2017	0.1864
0.0392	5.9209	14000	0.1985	0.1910
0.0338	6.3438	15000	0.1989	0.1807
0.0312	6.7668	16000	0.1982	0.1945
0.0237	7.1897	17000	0.1998	0.1842
0.0223	7.6126	18000	0.1994	0.1800
0.0192	8.0355	19000	0.1993	0.1806
0.0158	8.4584	20000	0.2000	0.1807

Framework versions

• Transformers 4.42.0.dev0
• Pytorch 2.1.0
• Datasets 2.19.1
• Tokenizers 0.19.1