wav2vec2-1b-Elderly5

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: 0.4667
• Cer: 12.2650

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: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• 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
8.8394	0.2580	200	2.3937	46.8691
1.5046	0.5160	400	1.3945	30.3689
1.0724	0.7741	600	1.0383	24.7180
0.8815	1.0321	800	0.9401	22.6797
0.7286	1.2901	1000	0.8012	20.3360
0.6375	1.5481	1200	0.6044	15.5016
0.5906	1.8062	1400	0.5698	14.4678
0.5084	2.0642	1600	0.5267	13.5280
0.4162	2.3222	1800	0.5424	13.9274
0.4088	2.5802	2000	0.4654	11.9596
0.3803	2.8383	2200	0.4667	12.2650

Framework versions

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