Gummybear05/wav2vec2-E30_freq_pause1

wav2vec2-E30_freq_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: 4.2019
• Cer: 75.5110

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
35.7962	0.1289	200	6.9049	100.0
14.8713	0.2579	400	4.9998	90.5193
10.3032	0.3868	600	4.5702	93.8440
8.67	0.5158	800	4.6552	92.9570
5.3475	0.6447	1000	4.5889	94.0437
4.9653	0.7737	1200	4.6626	93.9203
5.4367	0.9026	1400	4.4921	93.8322
5.1023	1.0316	1600	4.5898	93.5620
4.6675	1.1605	1800	4.3543	93.2801
4.9955	1.2895	2000	4.3195	92.9159
4.7843	1.4184	2200	4.2888	92.4871
4.7112	1.5474	2400	4.2545	92.2580
4.958	1.6763	2600	4.3499	85.8141
4.5195	1.8053	2800	4.2328	83.8522
4.7397	1.9342	3000	4.2644	82.5540
4.2707	2.0632	3200	4.3350	83.0298
4.6255	2.1921	3400	4.1961	81.6671
4.2181	2.3211	3600	4.1846	79.9107
4.6953	2.4500	3800	4.3664	74.8590
4.2375	2.5790	4000	4.5549	74.7650
4.186	2.7079	4200	4.2655	75.3818
4.2477	2.8369	4400	4.1909	76.5625
4.2201	2.9658	4600	4.2019	75.5110

Framework versions

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