wav2vec2-E30_speed_pause2

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: 0.8953
• Cer: 19.0834

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
30.708	0.1289	200	4.9281	100.0
4.9317	0.2579	400	4.6751	100.0
4.8066	0.3868	600	4.6333	100.0
4.7261	0.5158	800	4.5954	97.9142
4.6275	0.6447	1000	4.5410	97.9730
3.9356	0.7737	1200	3.3136	64.2362
2.7825	0.9026	1400	2.6449	46.5511
2.3364	1.0316	1600	2.2382	39.5770
2.0272	1.1605	1800	1.8952	34.2773
1.8032	1.2895	2000	1.8786	35.2291
1.6053	1.4184	2200	1.5099	27.7145
1.4602	1.5474	2400	1.4895	28.3373
1.3406	1.6763	2600	1.3510	27.1622
1.2236	1.8053	2800	1.2329	25.3819
1.1237	1.9342	3000	1.1621	23.7427
1.0466	2.0632	3200	1.1441	24.2127
0.9724	2.1921	3400	1.0863	22.8731
0.8808	2.3211	3600	1.0053	20.8696
0.86	2.4500	3800	0.9623	20.8343
0.8126	2.5790	4000	0.9202	20.0411
0.7842	2.7079	4200	0.9126	19.5182
0.7532	2.8369	4400	0.8995	19.2244
0.7269	2.9658	4600	0.8953	19.0834

Framework versions

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