wav2vec-malayalam-checkpoint

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6457
• Wer: 0.6608

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: 16
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 40
• num_epochs: 150
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.6371	10.0	100	3.5200	1.0
3.3014	20.0	200	3.2092	1.0
1.2997	30.0	300	0.7134	0.8847
0.5078	40.0	400	0.5805	0.7841
0.3795	50.0	500	0.5604	0.7289
0.2809	60.0	600	0.5962	0.7055
0.2381	70.0	700	0.6099	0.6938
0.2046	80.0	800	0.6237	0.6862
0.1826	90.0	900	0.6204	0.6755
0.1627	100.0	1000	0.6335	0.6751
0.1453	110.0	1100	0.6446	0.6739
0.1359	120.0	1200	0.6277	0.6648
0.1274	130.0	1300	0.6356	0.6573
0.1189	140.0	1400	0.6417	0.6601
0.1146	150.0	1500	0.6457	0.6608

Framework versions

• Transformers 4.11.3
• Pytorch 1.9.0+cu111
• Datasets 1.18.3
• Tokenizers 0.10.3