wav2vec2-Y_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: 1.7447
• Cer: 46.9631

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
28.6915	0.1289	200	4.9918	100.0
5.0255	0.2579	400	4.8286	100.0
4.8358	0.3868	600	4.6548	97.9676
4.7889	0.5158	800	4.6381	97.5681
4.687	0.6447	1000	4.5950	97.8560
4.5903	0.7737	1200	4.7109	96.0644
4.1984	0.9026	1400	4.4163	82.1546
3.4094	1.0316	1600	3.5488	77.3849
2.8656	1.1605	1800	3.2281	69.0202
2.5394	1.2895	2000	2.7503	63.1696
2.2135	1.4184	2200	2.6424	61.4603
2.0227	1.5474	2400	2.3955	57.3778
1.8118	1.6763	2600	2.2714	56.1913
1.7151	1.8053	2800	2.1585	52.1029
1.542	1.9342	3000	2.0588	51.3452
1.4724	2.0632	3200	1.9186	48.5667
1.3402	2.1921	3400	1.9230	50.2350
1.2483	2.3211	3600	1.8984	50.1997
1.2145	2.4500	3800	1.8578	48.7430
1.1362	2.5790	4000	1.7875	47.4800
1.1212	2.7079	4200	1.7618	47.7679
1.0894	2.8369	4400	1.7662	47.4977
1.0907	2.9658	4600	1.7447	46.9631

Framework versions

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