wav2vec2_xlsr50k_english_phoneme

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

• Loss: 0.5783
• Cer: 0.1178

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: 32
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
4.8403	6.94	500	1.1345	0.4657
0.5795	13.88	1000	0.3579	0.1169
0.3567	20.83	1500	0.3866	0.1174
0.2717	27.77	2000	0.4219	0.1169
0.2135	34.72	2500	0.4861	0.1199
0.1664	41.66	3000	0.5490	0.1179
0.1375	48.61	3500	0.5783	0.1178

Framework versions

• Transformers 4.22.0.dev0
• Pytorch 1.12.1
• Datasets 1.13.3
• Tokenizers 0.12.1