wav2vec2-large-xslr-commonvoice_jsut

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.7420
• Cer: 0.2028

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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: 1000
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
5.7253	0.6785	400	5.6492	0.9999
3.9768	1.3571	800	4.1057	0.9999
2.5285	2.0356	1200	2.3752	0.5055
1.0454	2.7142	1600	1.0980	0.2638
0.899	3.3927	2000	0.9496	0.2414
0.7883	4.0712	2400	0.8826	0.2310
0.7827	4.7498	2800	0.8434	0.2251
0.948	5.4283	3200	0.8063	0.2185
0.6689	6.1069	3600	0.7827	0.2142
0.5936	6.7854	4000	0.7687	0.2090
0.6368	7.4640	4400	0.7568	0.2065
0.5886	8.1425	4800	0.7554	0.2053
0.5633	8.8210	5200	0.7444	0.2036
0.6811	9.4996	5600	0.7420	0.2028

Framework versions

• Transformers 4.46.3
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.20.3