whisper-large-adalora-r8-st20k

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

• Loss: 0.5105

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• 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: 1000
• training_steps: 20000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.7557	0.3047	500	1.7281
0.7204	0.6094	1000	0.7339
0.6649	0.9141	1500	0.6413
0.6323	1.2188	2000	0.6106
0.5814	1.5235	2500	0.5928
0.6204	1.8282	3000	0.5797
0.5856	2.1328	3500	0.5699
0.5416	2.4375	4000	0.5621
0.5685	2.7422	4500	0.5560
0.588	3.0469	5000	0.5503
0.5546	3.3516	5500	0.5461
0.5448	3.6563	6000	0.5421
0.5577	3.9610	6500	0.5384
0.5568	4.2657	7000	0.5354
0.4886	4.5704	7500	0.5325
0.5432	4.8751	8000	0.5296
0.5107	5.1798	8500	0.5277
0.5114	5.4845	9000	0.5254
0.5063	5.7892	9500	0.5238
0.5207	6.0938	10000	0.5219
0.5193	6.3985	10500	0.5208
0.5257	6.7032	11000	0.5193
0.5132	7.0079	11500	0.5180
0.5206	7.3126	12000	0.5172
0.4857	7.6173	12500	0.5163
0.4775	7.9220	13000	0.5153
0.5069	8.2267	13500	0.5147
0.5079	8.5314	14000	0.5141
0.5231	8.8361	14500	0.5132
0.4779	9.1408	15000	0.5129
0.5314	9.4455	15500	0.5125
0.5046	9.7502	16000	0.5121
0.5483	10.0548	16500	0.5116
0.4921	10.3595	17000	0.5114
0.4573	10.6642	17500	0.5112
0.5476	10.9689	18000	0.5109
0.4629	11.2736	18500	0.5107
0.5538	11.5783	19000	0.5106
0.4906	11.8830	19500	0.5105
0.503	12.1877	20000	0.5105

Framework versions

• PEFT 0.15.2
• Transformers 4.52.3
• Pytorch 2.7.0+cu118
• Datasets 3.6.0
• Tokenizers 0.21.1