wav2vec2-xls-r-300m-fr-1h

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: 0.5292
• Wer: 0.3512
• Cer: 0.1108

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: 7e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.98) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.0018	11.3636	500	2.9956	1.0	1.0
2.879	22.7273	1000	2.9314	0.9958	0.9427
1.2342	34.0909	1500	0.8271	0.6077	0.1944
0.8487	45.4545	2000	0.5908	0.4622	0.1393
0.7073	56.8182	2500	0.5337	0.4119	0.1257
0.5277	68.1818	3000	0.5089	0.3837	0.1171
0.4473	79.5455	3500	0.5158	0.3666	0.1120
0.4438	90.9091	4000	0.5199	0.3610	0.1122
0.4289	102.2727	4500	0.5246	0.3496	0.1106
0.4218	113.6364	5000	0.5292	0.3512	0.1108

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.4.1
• Tokenizers 0.21.1