wav2vec2-large-xls-r-300m-vietnamese-cv11.0-colab

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6392
• Wer: 0.4792

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: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
10.0365	4.55	400	3.4508	0.9984
2.5036	9.09	800	1.0268	0.6972
0.5974	13.64	1200	0.7071	0.5492
0.3221	18.18	1600	0.6401	0.5071
0.2046	22.73	2000	0.6154	0.4871
0.1445	27.27	2400	0.6392	0.4792

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.0+cu113
• Datasets 2.8.0
• Tokenizers 0.10.3