wav2vec2-large-xls-r-300m-j-phoneme-common-test

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_10_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0000
• Wer: 0.0001

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1488	7.14	2000	0.0788	0.0919
0.0308	14.28	4000	0.0155	0.0271
0.0121	21.43	6000	0.0070	0.0103
0.0067	28.57	8000	0.0059	0.0067
0.0025	35.71	10000	0.0143	0.0180
0.0001	42.85	12000	0.0000	0.0001
0.0	50.0	14000	0.0000	0.0001

Framework versions

• Transformers 4.22.1
• Pytorch 1.10.0+cu113
• Datasets 2.5.1
• Tokenizers 0.12.1