npallewela/whisper-base-dv

Whisper Base Dv - Nuwan

This model is a fine-tuned version of openai/whisper-base on the Common Voice 15 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4586
• Wer Ortho: 283.5985
• Wer: 130.5086

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: 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: constant_with_warmup
• lr_scheduler_warmup_steps: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
1.0042	2.9586	500	1.0171	193.9489	109.2680
0.2061	5.9172	1000	0.2746	335.6644	168.7535
0.0655	8.8757	1500	0.2617	346.1823	165.9983
0.0202	11.8343	2000	0.3153	340.0754	150.7641
0.0073	14.7929	2500	0.3726	340.5918	142.2219
0.0059	17.7515	3000	0.4026	330.1717	147.8262
0.0034	20.7101	3500	0.4314	309.3663	137.8916
0.0026	23.6686	4000	0.4586	283.5985	130.5086

Framework versions

• Transformers 4.57.1
• Pytorch 2.8.0+cu126
• Datasets 3.6.0
• Tokenizers 0.22.1