wav2vec2-malayalam-binary

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1225
• Accuracy: 0.9659
• F1: 0.9677

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: 3e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.671	1.0	23	0.6930	0.5455	0.7059
0.6701	2.0	46	0.6928	0.5455	0.7059
0.6663	3.0	69	0.6922	0.5114	0.1887
0.6396	4.0	92	0.6185	0.9432	0.9462
0.5025	5.0	115	0.3883	0.9205	0.9278
0.34	6.0	138	0.2244	0.9318	0.9388
0.2135	7.0	161	0.1514	0.9545	0.9565
0.1553	8.0	184	0.0885	0.9886	0.9895
0.1271	9.0	207	0.0599	0.9886	0.9895
0.1021	10.0	230	0.1514	0.9545	0.9565
0.1047	11.0	253	0.0882	0.9773	0.9787
0.0649	12.0	276	0.1436	0.9545	0.9565
0.0976	13.0	299	0.1442	0.9545	0.9565
0.0535	14.0	322	0.1696	0.9545	0.9565
0.0431	15.0	345	0.0886	0.9659	0.9677
0.0817	16.0	368	0.1970	0.9545	0.9565
0.0463	17.0	391	0.2021	0.9545	0.9565
0.0573	18.0	414	0.1203	0.9659	0.9677
0.0346	19.0	437	0.1030	0.9659	0.9677
0.0493	20.0	460	0.1225	0.9659	0.9677

Framework versions

• Transformers 4.53.3
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.21.2