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: 1.0142
• Cer: 29.0179

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
29.2433	0.1289	200	4.9344	100.0
4.8669	0.2579	400	4.6589	100.0
4.7421	0.3868	600	4.6241	98.4258
4.6589	0.5158	800	4.5763	97.6269
4.5978	0.6447	1000	4.6357	97.5446
4.5014	0.7737	1200	4.3404	95.6884
3.8915	0.9026	1400	3.3848	65.5193
2.9112	1.0316	1600	2.6761	58.4763
2.4372	1.1605	1800	2.3130	48.8369
2.1033	1.2895	2000	2.0994	47.3391
1.8883	1.4184	2200	1.7943	42.6574
1.7159	1.5474	2400	1.7197	41.7998
1.5891	1.6763	2600	1.5193	38.6631
1.443	1.8053	2800	1.4734	37.8994
1.3674	1.9342	3000	1.3481	34.1988
1.2286	2.0632	3200	1.2702	32.8301
1.1511	2.1921	3400	1.2076	33.0592
1.0793	2.3211	3600	1.1945	33.2178
1.0503	2.4500	3800	1.1009	30.6039
1.016	2.5790	4000	1.0903	30.3102
0.988	2.7079	4200	1.0930	31.2441
0.957	2.8369	4400	1.0501	30.1692
0.9308	2.9658	4600	1.0142	29.0179

Framework versions

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