wav2vec2_Indonesia_4

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

• Loss: 1.3147
• Wer: 0.5914

Model description

Model yang dilatih dengan data train common voice dan data test data perkuliahan

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: 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
3.9949	3.23	400	1.3340	0.8916
0.4469	6.45	800	1.0507	0.6859
0.2003	9.68	1200	1.1115	0.6369
0.1432	12.9	1600	1.1307	0.6297
0.1138	16.13	2000	1.2157	0.6369
0.089	19.35	2400	1.2834	0.6058
0.0712	22.58	2800	1.3283	0.5947
0.057	25.81	3200	1.3345	0.5827
0.0467	29.03	3600	1.3147	0.5914

Framework versions

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