whisper-small-amh-matewosx

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

• Loss: 0.1151
• Wer: 0.4966
• Cer: 0.3684

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.4046	0.4312	500	0.1753	0.6207	0.4243
0.3112	0.8624	1000	0.1247	0.5390	0.3838
0.2479	1.2932	1500	0.1140	0.5178	0.3754
0.2359	1.7245	2000	0.1096	0.5081	0.3697
0.1810	2.1552	2500	0.1057	0.5026	0.3678
0.1960	2.5865	3000	0.1043	0.4983	0.3659
0.1504	3.0172	3500	0.1034	0.4944	0.3657
0.1576	3.4485	4000	0.1018	0.4947	0.3657
0.1560	3.8797	4500	0.0998	0.4908	0.3644
0.1191	4.3105	5000	0.1066	0.4940	0.3662
0.1231	4.7417	5500	0.1033	0.4900	0.3649
0.0932	5.1725	6000	0.1151	0.4966	0.3684

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2