wav2vec2-E50_freq_

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: 0.8668
• Cer: 19.9765

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
35.2248	0.1289	200	5.2165	100.0
4.8505	0.2579	400	4.6730	100.0
4.7409	0.3868	600	4.6010	100.0
4.7179	0.5158	800	4.5777	100.0
4.6385	0.6447	1000	4.5493	100.0
4.474	0.7737	1200	4.3678	95.6287
3.613	0.9026	1400	3.0539	52.1915
2.673	1.0316	1600	2.3161	41.9976
2.296	1.1605	1800	1.9627	37.4031
2.0153	1.2895	2000	1.8249	34.2597
1.8806	1.4184	2200	1.6718	32.1445
1.7151	1.5474	2400	1.4662	30.1175
1.5519	1.6763	2600	1.4178	29.8472
1.4466	1.8053	2800	1.2471	26.9330
1.3414	1.9342	3000	1.1782	24.7885
1.2326	2.0632	3200	1.1008	23.8425
1.1417	2.1921	3400	1.1546	25.1586
1.0752	2.3211	3600	1.0035	22.0035
1.0329	2.4500	3800	0.9728	21.5394
1.0197	2.5790	4000	0.9201	20.8284
0.9868	2.7079	4200	0.9076	20.2350
0.944	2.8369	4400	0.8786	19.8120
0.9381	2.9658	4600	0.8668	19.9765

Framework versions

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