w2v-bert-2.0-yoruba-CV17.0

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1095
• Wer: 0.1065

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 6
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3812	0.5102	100	0.3328	0.3070
0.2283	1.0204	200	0.2721	0.2807
0.1993	1.5306	300	0.3371	0.3481
0.2045	2.0408	400	0.3514	0.3314
0.2057	2.5510	500	0.3036	0.3086
0.2193	3.0612	600	0.2904	0.2847
0.1956	3.5714	700	0.2631	0.2534
0.1717	4.0816	800	0.1923	0.1995
0.1234	4.5918	900	0.1678	0.1732
0.0995	5.1020	1000	0.1280	0.1341
0.0614	5.6122	1100	0.1095	0.1065

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1