l3-whisper-small-l3c2_e4_v2

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

• Loss: 0.4322
• Wer: 45.6089

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2452	0.3230	10000	0.4564	42.0569
0.2149	0.6460	20000	0.4070	35.1088
0.196	0.9689	30000	0.3985	44.2632
0.1723	1.2919	40000	0.4032	38.9653
0.1663	1.6149	50000	0.4022	40.1084
0.163	1.9379	60000	0.4025	42.4103
0.1398	2.2608	70000	0.4187	43.5082
0.1401	2.5838	80000	0.4089	41.2552
0.1405	2.9068	90000	0.4187	46.2268
0.1173	3.2298	100000	0.4276	43.3936
0.1203	3.5527	110000	0.4309	43.3092
0.1167	3.8757	120000	0.4322	45.6089

Framework versions

• Transformers 4.51.3
• Pytorch 2.2.1
• Datasets 3.6.0
• Tokenizers 0.21.2