whisper-large-v3-turbo-ft-cv-cy

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the DewiBrynJones/commonvoice_18_0_cy train main dataset. It achieves the following results on the evaluation set:

• Loss: 0.4998
• Wer: 0.2181

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
0.1125	4.0161	1000	0.3743	0.2641
0.0178	8.0321	2000	0.4231	0.2499
0.0064	12.0482	3000	0.4627	0.2341
0.0004	16.0643	4000	0.4913	0.2201
0.0003	20.0803	5000	0.4998	0.2181

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1