6e-5_4000evaltest

This model was trained from scratch on the ami dataset. It achieves the following results on the evaluation set:

• Loss: 1.1036
• Wer: 0.2644

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: 6e-05
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	7.5758	250	6.5111	1.0
12.7243	15.1515	500	3.3744	1.0
12.7243	22.7273	750	3.0806	1.0
3.3334	30.3030	1000	3.2656	1.0
3.3334	37.8788	1250	2.9244	1.0
2.5277	45.4545	1500	1.2766	0.4895
2.5277	53.0303	1750	1.3037	0.3171
1.0439	60.6061	2000	1.0592	0.2950
1.0439	68.1818	2250	1.0711	0.2896
0.7466	75.7576	2500	1.0562	0.2809
0.7466	83.3333	2750	1.0679	0.2822
0.5899	90.9091	3000	1.1271	0.2734
0.5899	98.4848	3250	1.1434	0.2692
0.4165	106.0606	3500	1.0480	0.2683
0.4165	113.6364	3750	1.1392	0.2651
0.493	121.2121	4000	1.1036	0.2644

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1