whisper-large-v3-ft-btb-cv-ca-cy-2502

This model is a fine-tuned version of openai/whisper-large-v3 on the DewiBrynJones/banc-trawsgrifiadau-bangor-clean train main, DewiBrynJones/commonvoice_18_0_cy train+dev+other_with_excluded main, cymen-arfor/lleisiau-arfor train+dev main dataset. It achieves the following results on the evaluation set:

• Loss: 0.3689
• Wer: 0.2795

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5037	0.3638	1000	0.5148	0.3564
0.4137	0.7277	2000	0.4329	0.3181
0.282	1.0913	3000	0.4000	0.2959
0.2728	1.4552	4000	0.3815	0.2898
0.2743	1.8190	5000	0.3689	0.2795

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0