w2v-bert-cv-grain-lg_cv_only

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: inf
• Wer: 0.5800
• Cer: 0.1379

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
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.5013	1.0	2221	inf	0.2789	0.0724
0.299	2.0	4442	inf	0.2501	0.0648
0.2554	3.0	6663	inf	0.2435	0.0685
0.2411	4.0	8884	inf	0.2447	0.0648
0.2886	5.0	11105	inf	0.2506	0.0654
0.3923	6.0	13326	inf	0.4237	0.1108
2.1779	7.0	15547	inf	0.5612	0.1439
4.5629	8.0	17768	inf	0.5152	0.1379
2.236	9.0	19989	inf	0.5787	0.1384
2.2033	10.0	22210	inf	0.5742	0.1375
2.2047	11.0	24431	inf	0.5784	0.1382
2.2057	12.0	26652	inf	0.5805	0.1390
2.2076	13.0	28873	inf	0.5800	0.1379

Framework versions

• Transformers 4.46.1
• Pytorch 2.1.0+cu118
• Datasets 3.1.0
• Tokenizers 0.20.1