wav2vec2-large-xslr-commonvoice_jsut_split

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: 0.5074
• Cer: 0.1392

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
9.2478	0.8117	500	4.3917	0.9903
8.0651	1.6234	1000	4.0166	0.9903
2.7888	2.4351	1500	1.0882	0.2423
2.7409	3.2468	2000	0.7290	0.1821
1.6139	4.0584	2500	0.6303	0.1644
1.6685	4.8701	3000	0.5868	0.1561
1.4885	5.6818	3500	0.5589	0.1505
1.8946	6.4935	4000	0.5399	0.1470
1.3098	7.3052	4500	0.5269	0.1431
1.2762	8.1169	5000	0.5158	0.1409
1.2075	8.9286	5500	0.5122	0.1402
1.5473	9.7403	6000	0.5074	0.1392

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0