wav2vec2-ft-datastf5-1lre6-adm-ga2b16-st30k-pat3

This model is a fine-tuned version of jonatasgrosman/wav2vec2-large-english on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.0567
• Wer: 0.9996
• Cer: 0.9865

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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: 1500
• training_steps: 30000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
25.6288	0.6200	500	21.6325	1.8808	1.4695
5.6781	1.2393	1000	3.9369	1.0045	0.9887
3.8734	1.8593	1500	3.3859	0.9998	0.9861
3.407	2.4786	2000	3.3293	0.9996	0.9864
3.4348	3.0980	2500	3.2041	0.9996	0.9865
3.2992	3.7179	3000	3.1785	0.9996	0.9865
3.2306	4.3373	3500	3.1569	0.9996	0.9865
3.151	4.9572	4000	3.1325	0.9996	0.9865
3.1182	5.5766	4500	3.1203	0.9996	0.9865
3.1398	6.1959	5000	3.1219	0.9996	0.9865
3.1191	6.8159	5500	3.1054	0.9996	0.9865
3.1332	7.4352	6000	3.0948	0.9996	0.9865
3.0845	8.0546	6500	3.0861	0.9996	0.9865
3.0783	8.6745	7000	3.0861	0.9996	0.9865
3.0583	9.2939	7500	3.0810	0.9996	0.9865
3.1264	9.9138	8000	3.0866	0.9996	0.9865
3.0891	10.5332	8500	3.0841	0.9996	0.9865
3.0803	11.1525	9000	3.0768	0.9996	0.9865
3.1139	11.7725	9500	3.0600	0.9996	0.9865
3.0704	12.3918	10000	3.0682	0.9996	0.9865
2.9783	13.0112	10500	3.0559	0.9996	0.9865
3.0563	13.6311	11000	3.0641	0.9996	0.9865
3.0726	14.2505	11500	3.0537	0.9996	0.9865
3.0792	14.8704	12000	3.0560	0.9996	0.9865
3.0304	15.4898	12500	3.0579	0.9996	0.9865
3.0168	16.1091	13000	3.0567	0.9996	0.9865

Framework versions

• Transformers 4.52.3
• Pytorch 2.7.0+cu118
• Datasets 3.6.0
• Tokenizers 0.21.1