wav2vec2-1b-E3

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4214
• Cer: 11.5484

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: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• 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
10.9096	0.2580	200	4.4784	97.7502
2.5397	0.5160	400	1.5540	34.0754
1.2018	0.7741	600	1.1079	26.2805
0.9714	1.0321	800	0.8312	20.3360
0.7789	1.2901	1000	0.7358	18.0980
0.6942	1.5481	1200	0.6558	17.2580
0.6176	1.8062	1400	0.5847	15.0787
0.5476	2.0642	1600	0.5941	15.8365
0.4525	2.3222	1800	0.5006	13.3870
0.4068	2.5802	2000	0.4577	12.6527
0.37	2.8383	2200	0.4214	11.5484

Framework versions

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