whisper-tiny-en-atc-thesis-2-no-lora

This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.7688
• Wer: 44.8980

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: 1
• 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: 500
• training_steps: 15000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4276	250.0	500	0.2941	22.4490
0.0001	500.0	1000	0.4339	24.4898
0.0	750.0	1500	0.5702	24.4898
0.0	1000.0	2000	0.7041	28.5714
0.0	1250.0	2500	0.7371	28.5714
0.0	1500.0	3000	0.8153	32.6531
0.0	1750.0	3500	0.8885	26.5306
0.0	2000.0	4000	0.9523	24.4898
0.0	2250.0	4500	0.9644	38.7755
0.0	2500.0	5000	1.0169	32.6531
0.0	2750.0	5500	1.0098	34.6939
0.0	3000.0	6000	1.0696	32.6531
0.0435	3250.0	6500	0.6549	26.5306
0.0	3500.0	7000	0.8819	28.5714
0.0	3750.0	7500	1.0423	30.6122
0.0	4000.0	8000	1.2150	32.6531
0.0	4250.0	8500	1.3003	32.6531
0.0	4500.0	9000	1.4076	36.7347
0.0	4750.0	9500	1.5208	38.7755
0.0	5000.0	10000	1.6303	38.7755
0.0	5250.0	10500	1.6312	38.7755
0.0	5500.0	11000	1.6982	38.7755
0.0	5750.0	11500	1.7714	42.8571
0.0	6000.0	12000	1.8436	42.8571
0.0	6250.0	12500	1.7950	44.8980
0.0	6500.0	13000	1.8284	44.8980
0.0	6750.0	13500	1.8639	44.8980
0.0	7000.0	14000	1.8944	44.8980
0.0	7250.0	14500	1.7909	44.8980
0.0	7500.0	15000	1.7688	44.8980

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3