nectar_aihub_model_15000steps

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

• Loss: 0.1385
• Cer: 11.6172

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: 8
• 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: 15000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.225	0.1307	1500	0.2025	14.2277
0.1842	0.2614	3000	0.1862	12.2103
0.1767	0.3922	4500	0.1746	11.7624
0.1459	0.5229	6000	0.1656	11.3708
0.15	0.6536	7500	0.1577	11.1017
0.1415	0.7843	9000	0.1517	11.3546
0.1559	0.9150	10500	0.1446	10.7843
0.0772	1.0458	12000	0.1430	10.9494
0.073	1.1765	13500	0.1405	11.8585
0.0737	1.3072	15000	0.1385	11.6172

Framework versions

• Transformers 4.41.2
• Pytorch 2.2.2+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1