wav2vec2-E_

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.7138
• Cer: 15.1116

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: 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: 50
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
29.9637	0.1289	200	4.9791	100.0
4.6123	0.2579	400	4.6053	100.0
4.546	0.3868	600	4.5594	97.9671
4.49	0.5158	800	4.5282	97.9671
4.4469	0.6447	1000	4.4909	97.9671
4.3231	0.7737	1200	4.3157	89.9412
3.7783	0.9026	1400	3.3710	58.4195
2.7113	1.0316	1600	2.4292	42.7321
2.0971	1.1605	1800	2.0556	38.2197
1.7684	1.2895	2000	1.6070	29.8472
1.5046	1.4184	2200	1.4542	27.2268
1.3672	1.5474	2400	1.2693	24.2656
1.2279	1.6763	2600	1.1761	23.5253
1.1456	1.8053	2800	1.0789	21.2338
1.0516	1.9342	3000	0.9792	19.9471
0.9609	2.0632	3200	0.9519	19.6592
0.8758	2.1921	3400	0.8677	18.0082
0.8289	2.3211	3600	0.8783	18.0082
0.783	2.4500	3800	0.7990	17.0682
0.7719	2.5790	4000	0.7492	15.7168
0.748	2.7079	4200	0.7483	16.0400
0.7377	2.8369	4400	0.7291	15.4877
0.7188	2.9658	4600	0.7138	15.1116

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3