wav2vec2-E10_pause

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: 1.2932
• Cer: 28.7124

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
27.1594	0.1289	200	4.8622	100.0
4.9998	0.2579	400	4.7518	100.0
4.8719	0.3868	600	4.7669	100.0
4.809	0.5158	800	4.6593	100.0
4.7354	0.6447	1000	4.5911	100.0
4.6679	0.7737	1200	4.6423	99.3773
4.136	0.9026	1400	3.9276	77.9077
3.1108	1.0316	1600	2.9616	56.7845
2.6314	1.1605	1800	2.7039	51.9619
2.2786	1.2895	2000	2.3306	45.8823
2.0348	1.4184	2200	2.1354	40.5721
1.8952	1.5474	2400	1.9727	39.7086
1.7053	1.6763	2600	1.8535	37.7996
1.5809	1.8053	2800	1.7608	36.7246
1.4968	1.9342	3000	1.6229	33.2531
1.349	2.0632	3200	1.6171	33.6290
1.2592	2.1921	3400	1.5156	32.9300
1.2043	2.3211	3600	1.4406	30.7977
1.1418	2.4500	3800	1.3878	29.5172
1.1157	2.5790	4000	1.3441	29.1060
1.0653	2.7079	4200	1.3052	27.9605
1.0451	2.8369	4400	1.2943	28.5656
1.0225	2.9658	4600	1.2932	28.7124

Framework versions

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