wav2vec2-large-xls-r-300m-vi-75p

This model is a fine-tuned version of leviethoang/wav2vec2-large-xls-r-300m-vi-25p on the common_voice dataset. It achieves the following results on the evaluation set:

• Loss: 1.7880
• Wer: 0.4324

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• 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: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.962	1.68	400	1.2033	0.4428
0.7977	3.36	800	1.3410	0.4731
0.644	5.04	1200	1.4682	0.4796
0.5156	6.72	1600	1.4940	0.4826
0.4531	8.4	2000	1.5071	0.4734
0.3882	10.08	2400	1.5408	0.4694
0.3469	11.76	2800	1.5975	0.4697
0.3096	13.45	3200	1.7120	0.4728
0.2825	15.13	3600	1.7052	0.4632
0.2607	16.81	4000	1.6870	0.4575
0.2301	18.49	4400	1.7205	0.4653
0.2096	20.17	4800	1.7352	0.4504
0.1915	21.85	5200	1.7948	0.4465
0.1685	23.53	5600	1.7994	0.4400
0.1543	25.21	6000	1.7613	0.4435
0.1378	26.89	6400	1.8300	0.4365
0.1278	28.57	6800	1.7880	0.4324

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.0+cu113
• Datasets 1.18.3
• Tokenizers 0.10.3