wav2vec2-E30_fsp_

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: 1.2157
• Cer: 25.9927

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
32.1516	0.1289	200	5.0943	100.0
4.9521	0.2579	400	4.7057	100.0
4.8124	0.3868	600	4.6675	100.0
4.7767	0.5158	800	4.6317	100.0
4.6927	0.6447	1000	4.5729	100.0
4.6586	0.7737	1200	4.5402	97.9676
4.5988	0.9026	1400	4.4582	97.5505
4.4156	1.0316	1600	4.2520	80.2220
3.7038	1.1605	1800	3.2628	60.0681
2.9837	1.2895	2000	2.8047	52.5611
2.5915	1.4184	2200	2.4365	44.0555
2.2905	1.5474	2400	2.1633	41.8174
2.0879	1.6763	2600	1.9046	38.5515
1.8934	1.8053	2800	1.7398	34.6511
1.7379	1.9342	3000	1.6517	32.7655
1.5844	2.0632	3200	1.5588	32.0724
1.4751	2.1921	3400	1.4593	30.0517
1.3796	2.3211	3600	1.4139	29.4055
1.3051	2.4500	3800	1.3397	28.1720
1.2762	2.5790	4000	1.2967	27.3144
1.2411	2.7079	4200	1.2657	26.7563
1.1701	2.8369	4400	1.2403	26.3980
1.1411	2.9658	4600	1.2157	25.9927

Framework versions

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