abbenedekwhisper-base.en-finetuned-D3K

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

• Loss: 1.5119
• Cer: 21.9254
• Wer: 27.8146
• Ser: 16.6667

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-06
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 40
• training_steps: 400
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer	Ser
7.7093	0.21	20	6.8991	20.3673	29.4702	28.9474
6.8825	0.43	40	5.5765	25.0417	35.0993	30.7018
5.2019	0.64	60	4.4037	30.4953	41.7219	34.2105
3.8746	0.85	80	3.3761	33.5559	45.0331	32.4561
2.9591	1.06	100	2.7073	32.1091	43.0464	30.7018
2.2502	1.28	120	2.2711	33.3890	44.3709	30.7018
1.8478	1.49	140	1.9790	31.4413	41.0596	27.1930
1.6338	1.7	160	1.8207	32.2204	42.0530	26.3158
1.5065	1.91	180	1.7257	26.4329	34.4371	20.1754
1.3667	2.13	200	1.6645	24.5965	31.4570	19.2982
1.2807	2.34	220	1.6155	24.8191	31.7881	20.1754
1.2212	2.55	240	1.5851	26.2104	33.4437	21.0526
1.2299	2.77	260	1.5669	24.4853	31.4570	19.2982
1.1623	2.98	280	1.5543	23.0940	29.8013	18.4211
1.1275	3.19	300	1.5421	23.0940	29.8013	18.4211
1.1304	3.4	320	1.5320	23.5392	29.8013	18.4211
1.1113	3.62	340	1.5236	21.9254	27.8146	16.6667
1.1189	3.83	360	1.5168	21.9254	27.8146	16.6667
1.1134	4.04	380	1.5134	21.9254	27.8146	16.6667
1.097	4.26	400	1.5119	21.9254	27.8146	16.6667

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.2+cu121
• Datasets 2.14.5
• Tokenizers 0.15.2