wav2vec2-E50_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: 1.3869
• Cer: 26.1751

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
29.7901	0.1289	200	4.9226	100.0
4.8707	0.2579	400	4.6818	100.0
4.782	0.3868	600	4.6430	100.0
4.7305	0.5158	800	4.6881	100.0
4.6769	0.6447	1000	4.6334	100.0
4.6788	0.7737	1200	4.5911	100.0
4.6287	0.9026	1400	4.5118	100.0
4.5731	1.0316	1600	4.6535	100.0
4.541	1.1605	1800	4.9000	100.0
4.4683	1.2895	2000	4.3794	98.6134
4.1608	1.4184	2200	3.8468	70.1351
3.2361	1.5474	2400	2.9225	51.1927
2.6741	1.6763	2600	2.4824	43.5311
2.3514	1.8053	2800	2.1972	39.7767
2.1265	1.9342	3000	1.9765	35.1293
1.9561	2.0632	3200	1.8913	32.9671
1.814	2.1921	3400	1.7420	31.9330
1.663	2.3211	3600	1.6148	29.0717
1.5849	2.4500	3800	1.5338	28.2550
1.5459	2.5790	4000	1.5955	29.9530
1.4935	2.7079	4200	1.4618	27.8496
1.414	2.8369	4400	1.4365	26.9389
1.3898	2.9658	4600	1.3869	26.1751

Framework versions

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