klem-whisper-large-v3

This model is a fine-tuned version of openai/whisper-medium on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3933
• Wer: 31.3665

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAFACTOR and the args are: No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
11.0127	0.2211	500	0.6602	48.0645
8.8798	0.4423	1000	0.5308	41.5296
7.7174	0.6634	1500	0.4741	40.0829
7.7663	0.8845	2000	0.4346	35.4447
5.1383	1.1053	2500	0.4138	34.0724
4.7879	1.3264	3000	0.4084	33.0302
5.3680	1.5475	3500	0.4087	32.8101
5.0507	1.7686	4000	0.3982	32.454
4.8131	1.9898	4500	0.3906	31.8099
3.8230	2.2105	5000	0.3933	31.3665

Framework versions

• Transformers 5.6.2
• Pytorch 2.10.0+cu128
• Datasets 4.8.4
• Tokenizers 0.22.2