wav2vec2_ljspeech_with_stress

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0431
• Cer: 0.0073

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: 0.0001
• 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: 1000
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
2.9043	1.53	500	3.0446	1.0
1.1195	3.05	1000	0.1302	0.0233
0.1656	4.58	1500	0.0728	0.0149
0.1136	6.11	2000	0.0581	0.0122
0.0852	7.63	2500	0.0508	0.0102
0.0746	9.16	3000	0.0472	0.0093
0.0646	10.69	3500	0.0443	0.0084
0.0588	12.21	4000	0.0442	0.0081
0.0513	13.74	4500	0.0437	0.0077
0.046	15.27	5000	0.0435	0.0075
0.0446	16.79	5500	0.0430	0.0075
0.0429	18.32	6000	0.0433	0.0074
0.0412	19.85	6500	0.0431	0.0072

Framework versions

• Transformers 4.34.0
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1