Model_G_2

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: 0.3710
• Wer: 0.2513
• Cer: 0.0631

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
3.7484	3.23	400	0.5706	0.5698	0.1477
0.3419	6.45	800	0.4120	0.3758	0.0924
0.1796	9.68	1200	0.3691	0.3295	0.0843
0.125	12.9	1600	0.3821	0.3097	0.0782
0.0984	16.13	2000	0.4085	0.2947	0.0742
0.0827	19.35	2400	0.3859	0.2781	0.0711
0.0666	22.58	2800	0.3813	0.2663	0.0684
0.0558	25.81	3200	0.3681	0.2545	0.0644
0.0466	29.03	3600	0.3710	0.2513	0.0631

Framework versions

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