wav2vec2-1b-Elderly3

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: 0.9748
• Cer: 23.8369

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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
10.1541	0.2580	200	4.8629	92.0935
4.4064	0.5160	400	4.6177	92.0524
4.2096	0.7739	600	4.1126	80.3278
3.6968	1.0319	800	3.1236	65.1081
2.6923	1.2899	1000	2.3085	48.5374
2.0244	1.5479	1200	1.8248	40.0435
1.6074	1.8059	1400	1.4809	33.3999
1.3326	2.0639	1600	1.2833	31.2383
1.0925	2.3218	1800	1.1624	28.2072
1.0068	2.5798	2000	1.1147	26.7681
0.8894	2.8378	2200	0.9748	23.8369

Framework versions

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