Tiberiw/whisper-BASE-LORA-med-merged

whisper-BASE-LORA-med

This model is a fine-tuned version of openai/whisper-base on the Shamus/multimed_short dataset. It achieves the following results on the evaluation set:

• Loss: 0.5848
• Wer: 21.4724
• Cer: 12.0240

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: 0.0005
• train_batch_size: 4
• 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
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.2829	1.0	7077	0.6521	24.4961	14.2032
0.8621	2.0	14154	0.6147	24.9126	14.5952
0.3164	3.0	21231	0.5794	22.2598	12.4891
0.3039	4.0	28308	0.5678	22.0960	12.5164
0.5835	5.0	35385	0.5848	21.4724	12.0240

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1