whisper-tiny-tedlium

This model is a fine-tuned version of openai/whisper-tiny.en on the tedlium dataset. It achieves the following results on the evaluation set:

• Loss: 2.2271
• Wer: 24.2451

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: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.06	20	3.5590	35.2108
3.8059	0.13	40	3.0962	34.2645
2.6288	0.19	60	2.7592	31.2517
1.688	0.26	80	2.4428	24.8330
1.2352	0.32	100	2.3773	22.4951
1.2352	0.38	120	2.4088	21.6219
0.9947	0.45	140	2.3161	19.5345
0.9169	0.51	160	2.2712	23.5736
0.8418	0.58	180	2.1986	23.2918
0.8253	0.64	200	2.2271	24.2451

Framework versions

• Transformers 4.39.2
• Pytorch 2.2.2+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2