w2v-bert-2.0-hausa_100_400h_yourtts

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the CLEAR-GLOBAL/HAUSA_100_400H_YOURTTS - NA dataset. It achieves the following results on the evaluation set:

• Loss: 0.2925
• Wer: 0.3666
• Cer: 0.1987

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: 3e-05
• train_batch_size: 160
• eval_batch_size: 160
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 320
• total_eval_batch_size: 320
• optimizer: Use OptimizerNames.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_ratio: 0.1
• num_epochs: 50.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.6098	0.8764	1000	0.4685	0.4367	0.2187
0.0454	1.7528	2000	0.3571	0.3927	0.2063
0.1228	2.6293	3000	0.3201	0.3700	0.2001
0.0257	3.5057	4000	0.2908	0.3666	0.1986
0.028	4.3821	5000	0.3164	0.3757	0.2010
0.1503	5.2585	6000	0.2917	0.3521	0.1955
0.0199	6.1350	7000	0.2963	0.3565	0.1962
0.0205	7.0114	8000	0.2909	0.3557	0.1968
0.0232	7.8878	9000	0.2975	0.3578	0.1963

Framework versions

• Transformers 4.48.1
• Pytorch 2.7.0+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1