whisper-small-bn

This model is a fine-tuned version of openai/whisper-small on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1134
• Wer: 31.7721

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 500
• training_steps: 6000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2262	0.3183	500	0.2355	61.4003
0.1542	0.6365	1000	0.1628	47.9933
0.1166	0.9548	1500	0.1391	43.0509
0.0803	1.2731	2000	0.1252	38.7278
0.0793	1.5913	2500	0.1147	36.4728
0.0834	1.9096	3000	0.1079	34.6947
0.0463	2.2279	3500	0.1097	33.4679
0.0541	2.5461	4000	0.1084	32.9231
0.0487	2.8644	4500	0.1042	32.5743
0.0281	3.1827	5000	0.1136	32.6657
0.0313	3.5010	5500	0.1127	32.2032
0.0286	3.8192	6000	0.1134	31.7721

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1
• Datasets 2.20.0
• Tokenizers 0.19.1