wavlm-tamil-binary

This model is a fine-tuned version of microsoft/wavlm-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7413
• Accuracy: 0.8039
• F1: 0.7059

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
0.6795	1.0	13	0.6894	0.7059	0.7059
0.665	2.0	26	0.6510	0.6863	0.4286
0.5559	3.0	39	0.4561	0.8235	0.8
0.4817	4.0	52	0.4987	0.8039	0.7368
0.4534	5.0	65	0.4739	0.7647	0.6842
0.4044	6.0	78	0.4715	0.8235	0.7805
0.3852	7.0	91	0.3756	0.8235	0.8
0.3803	8.0	104	0.4911	0.7843	0.7179
0.304	9.0	117	0.6025	0.7843	0.6667
0.2876	10.0	130	0.5875	0.7843	0.6667
0.2543	11.0	143	0.4856	0.8235	0.7568
0.2159	12.0	156	0.4566	0.8431	0.7895
0.1855	13.0	169	0.3982	0.8627	0.8205
0.1821	14.0	182	0.8304	0.7647	0.625
0.1803	15.0	195	0.8396	0.7647	0.625
0.1333	16.0	208	0.6040	0.8235	0.7429
0.1249	17.0	221	0.5015	0.8627	0.8205
0.1505	18.0	234	0.7057	0.8039	0.7059
0.1287	19.0	247	0.7512	0.8039	0.7059
0.1018	20.0	260	0.7413	0.8039	0.7059

Framework versions

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