wav2vec2-E30_feq_speed2

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: 1.1152
• Cer: 23.6251

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
40.1347	0.1289	200	5.0362	100.0
4.9594	0.2579	400	4.6347	100.0
4.7937	0.3868	600	4.6122	100.0
4.7775	0.5158	800	4.6742	100.0
4.7068	0.6447	1000	4.5784	100.0
4.6819	0.7737	1200	4.5673	100.0
4.6273	0.9026	1400	4.4307	100.0
4.355	1.0316	1600	4.0447	88.3020
3.6473	1.1605	1800	3.1079	59.8884
2.8791	1.2895	2000	2.5597	49.6651
2.4172	1.4184	2200	2.2464	44.2656
2.0866	1.5474	2400	1.9422	39.3948
1.8889	1.6763	2600	1.7683	34.9060
1.7055	1.8053	2800	1.6691	33.8719
1.5474	1.9342	3000	1.4660	29.6886
1.3932	2.0632	3200	1.4206	28.7427
1.2787	2.1921	3400	1.3387	27.7086
1.1968	2.3211	3600	1.2747	26.5217
1.1476	2.4500	3800	1.2341	25.7168
1.1007	2.5790	4000	1.1612	24.3831
1.0757	2.7079	4200	1.1585	24.6416
1.0166	2.8369	4400	1.1327	23.9953
0.9917	2.9658	4600	1.1152	23.6251

Framework versions

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