english-filipino-wav2vec2-l-xls-r-test-05

Model description

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

• Loss: 0.4738
• Wer: 0.2684

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
3.3328	2.09	400	2.2174	0.9733
0.6432	4.19	800	0.3735	0.3896
0.2741	6.28	1200	0.3639	0.3425
0.1877	8.38	1600	0.3506	0.3425
0.1408	10.47	2000	0.3644	0.3181
0.1133	12.57	2400	0.3837	0.3047
0.0953	14.66	2800	0.4415	0.3103
0.0814	16.75	3200	0.3940	0.3092
0.0707	18.85	3600	0.4164	0.3013
0.059	20.94	4000	0.4488	0.2983
0.0545	23.04	4400	0.4803	0.3028
0.0482	25.13	4800	0.4731	0.2811
0.0426	27.23	5200	0.4606	0.2757
0.0395	29.32	5600	0.4738	0.2684

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Framework versions

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