wav2vec2-Y_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: 1.6778
• Cer: 39.4267

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
5.1062	0.1290	200	4.7053	100.0
4.8751	0.2581	400	4.8829	100.0
4.7665	0.3871	600	4.6329	98.8781
4.6575	0.5161	800	4.7058	98.4199
4.2511	0.6452	1000	4.2469	90.7777
3.1636	0.7742	1200	3.3817	69.3844
2.6261	0.9032	1400	2.9457	60.4676
2.1994	1.0323	1600	2.6949	56.0092
1.924	1.1613	1800	2.5125	52.3085
1.7291	1.2903	2000	2.2571	49.5653
1.5934	1.4194	2200	2.0517	46.2523
1.5086	1.5484	2400	2.1590	46.3757
1.4041	1.6774	2600	2.0795	46.1407
1.3266	1.8065	2800	2.1936	47.5388
1.2494	1.9355	3000	2.0095	45.1891
1.1305	2.0645	3200	1.8807	43.5092
1.0493	2.1935	3400	1.7053	40.0141
0.9978	2.3226	3600	1.8685	43.1508
0.9689	2.4516	3800	1.8416	41.8938
0.9527	2.5806	4000	1.7686	42.1405
0.8927	2.7097	4200	1.7281	40.0611
0.8958	2.8387	4400	1.6940	39.6264
0.8855	2.9677	4600	1.6778	39.4267

Framework versions

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