sarpba/whisper-tiny-cv16.1-hu-v10-cleaned

Whisper Tiny Hu v10 - cleaned

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 16.1 hu cleaned dataset. It achieves the following results on the evaluation set:

• Loss: 0.2366
• Wer Ortho: 21.1511
• Wer: 19.9338

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: 4e-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: constant_with_warmup
• lr_scheduler_warmup_steps: 300
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.2738	0.83	1000	0.2861	32.5590	30.9570
0.1454	1.66	2000	0.2307	26.5776	25.1495
0.0811	2.49	3000	0.2144	24.4515	23.2845
0.0424	3.32	4000	0.2118	22.5739	21.3068
0.0248	4.15	5000	0.2154	22.3803	21.2875
0.0309	4.99	6000	0.2116	22.2384	21.0753
0.0254	5.82	7000	0.2211	22.2093	21.1493
0.0146	6.65	8000	0.2262	22.4545	21.3776
0.0134	7.48	9000	0.2294	21.2156	20.0238
0.011	8.31	10000	0.2366	21.1511	19.9338

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.0+cu121
• Datasets 2.17.0
• Tokenizers 0.15.1