wav2vec2-Y_freq_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: 1.2487
• Cer: 27.1563

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
45.9774	0.1289	200	5.3401	100.0
5.0614	0.2579	400	4.6734	100.0
4.8698	0.3868	600	4.6465	100.0
4.8405	0.5158	800	4.6147	100.0
4.7495	0.6447	1000	4.5812	100.0
4.6725	0.7737	1200	4.4809	100.0
4.4715	0.9026	1400	4.2162	98.0964
3.6966	1.0316	1600	3.0934	60.3114
2.8837	1.1605	1800	2.7554	51.6980
2.3994	1.2895	2000	2.3323	44.8766
2.1484	1.4184	2200	2.0952	41.4806
1.9221	1.5474	2400	1.9481	39.1951
1.8075	1.6763	2600	1.7922	35.6345
1.6334	1.8053	2800	1.7316	35.6404
1.5031	1.9342	3000	1.5918	33.4841
1.3591	2.0632	3200	1.4975	30.9401
1.2379	2.1921	3400	1.4342	30.5817
1.1751	2.3211	3600	1.3655	28.4900
1.1049	2.4500	3800	1.3425	28.6780
1.0825	2.5790	4000	1.2924	27.9377
1.0405	2.7079	4200	1.2683	27.2973
0.9975	2.8369	4400	1.2658	27.6204
0.9624	2.9658	4600	1.2487	27.1563

Framework versions

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