whisper-large-v3-LORA-uk

This model is a fine-tuned version of openai/whisper-large-v3 on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2198
• Wer: 32.3224
• Cer: 11.7944

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: 0.001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.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: 6000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.3078	0.4034	1000	0.2892	46.7412	17.3789
0.238	0.8068	2000	0.2869	47.2200	18.3277
0.1942	1.2102	3000	0.2419	44.1075	15.9300
0.1773	1.6136	4000	0.2262	30.5134	8.7628
0.0626	2.0169	5000	0.2192	34.1580	11.7469
0.0785	2.4203	6000	0.2198	32.3224	11.7944

Framework versions

• PEFT 0.15.2.dev0
• Transformers 4.53.0.dev0
• Pytorch 2.5.1+cu124
• Datasets 3.5.0
• Tokenizers 0.21.0