wav2vec2-1b-Elderly2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4163
• Cer: 11.6894

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: 0.0001
• train_batch_size: 8
• 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: 50
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
10.2843	0.1289	200	2.5492	45.7766
1.8359	0.2579	400	1.5587	34.1694
1.3781	0.3868	600	1.5687	34.6335
1.1717	0.5158	800	1.1218	26.3217
1.1155	0.6447	1000	1.0626	26.6095
1.018	0.7737	1200	0.9805	24.6828
0.9304	0.9026	1400	0.9428	23.6783
0.8523	1.0316	1600	0.8593	22.5094
0.7385	1.1605	1800	0.7612	20.2420
0.7027	1.2895	2000	0.7923	20.9410
0.702	1.4184	2200	0.6695	17.7867
0.6454	1.5474	2400	0.6731	18.1391
0.5988	1.6763	2600	0.6737	17.6281
0.5968	1.8053	2800	0.6260	16.2007
0.5786	1.9342	3000	0.6055	16.0068
0.4825	2.0632	3200	0.6158	16.3828
0.4486	2.1921	3400	0.5301	13.9626
0.4043	2.3211	3600	0.5068	14.0155
0.3979	2.4500	3800	0.5188	13.9215
0.3779	2.5790	4000	0.4591	12.7643
0.3663	2.7079	4200	0.4430	12.4413
0.3539	2.8369	4400	0.4293	11.9537
0.3436	2.9658	4600	0.4163	11.6894

Framework versions

• Transformers 4.45.2
• Pytorch 2.3.1.post100
• Datasets 2.19.1
• Tokenizers 0.20.1