whisper-medium-ft-btb-cv-cvad-ca-cy-2504

This model is a fine-tuned version of openai/whisper-medium on the DewiBrynJones/banc-trawsgrifiadau-bangor train main, DewiBrynJones/commonvoice_18_0_cy train+dev+other_with_excluded main, cymen-arfor/lleisiau-arfor train+dev main, techiaith/commonvoice_vad_cy train main dataset. It achieves the following results on the evaluation set:

• Loss: 0.5789
• Wer: 0.3069

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: 64
• eval_batch_size: 16
• seed: 42
• optimizer: Use adamw_torch 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
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5031	0.6116	1000	0.5204	0.3768
0.3161	1.2232	2000	0.4475	0.3410
0.3062	1.8349	3000	0.4146	0.3239
0.2002	2.4465	4000	0.4276	0.3040
0.1148	3.0581	5000	0.4366	0.3116
0.1114	3.6697	6000	0.4480	0.3012
0.0578	4.2813	7000	0.4946	0.3038
0.0539	4.8930	8000	0.5114	0.3041
0.0235	5.5046	9000	0.5628	0.3089
0.0159	6.1162	10000	0.5789	0.3069

Framework versions

• Transformers 4.51.3
• Pytorch 2.7.0+cu126
• Datasets 3.5.1
• Tokenizers 0.21.1