Model_G_Wav2Vec2_Version3

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice dataset. It achieves the following results on the evaluation set:

• Loss: 0.4863
• Wer: 0.3321
• Cer: 0.0851

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: 0.0003
• 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: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.7935	5.97	400	0.6443	0.6220	0.1605
0.2843	11.94	800	0.5294	0.4286	0.1090
0.1364	17.91	1200	0.4766	0.3774	0.0969
0.0914	23.88	1600	0.4960	0.3408	0.0880
0.0662	29.85	2000	0.4863	0.3321	0.0851

Framework versions

• Transformers 4.31.0
• Pytorch 2.0.1+cu117
• Datasets 1.18.3
• Tokenizers 0.13.3