wav2vec2-E30_speed

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: 1.5403
• Cer: 30.0576

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
38.4014	0.1289	200	5.0924	100.0
4.967	0.2579	400	4.7266	100.0
4.803	0.3868	600	4.6563	100.0
4.7976	0.5158	800	4.7264	100.0
4.7401	0.6447	1000	4.6214	100.0
4.7227	0.7737	1200	4.6575	100.0
4.6903	0.9026	1400	4.6785	100.0
4.5596	1.0316	1600	4.4755	100.0
4.3655	1.1605	1800	4.1150	90.4312
3.7719	1.2895	2000	3.5459	66.0186
3.2218	1.4184	2200	2.9444	54.4643
2.9251	1.5474	2400	2.7016	50.6344
2.6652	1.6763	2600	2.5111	45.5416
2.4444	1.8053	2800	2.2185	40.9246
2.2701	1.9342	3000	2.0419	38.7101
2.1017	2.0632	3200	1.9275	36.5836
1.982	2.1921	3400	1.9218	36.9302
1.8856	2.3211	3600	1.7415	33.2648
1.7977	2.4500	3800	1.7043	33.2942
1.7146	2.5790	4000	1.6084	30.8036
1.6798	2.7079	4200	1.5947	31.0033
1.6234	2.8369	4400	1.5576	30.2573
1.6289	2.9658	4600	1.5403	30.0576

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1