wavlm-telugu-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: 0.7316
• Accuracy: 0.8545
• F1: 0.8919

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
0.6782	1.0	14	0.6797	0.4727	0.3256
0.6535	2.0	28	0.5998	0.7818	0.8235
0.5762	3.0	42	0.4867	0.8	0.8358
0.4791	4.0	56	0.5116	0.8182	0.8611
0.426	5.0	70	0.5037	0.7818	0.8182
0.374	6.0	84	0.4857	0.8364	0.8696
0.3461	7.0	98	0.4698	0.8182	0.8529
0.3322	8.0	112	0.5055	0.8364	0.88
0.201	9.0	126	0.5055	0.8545	0.8919
0.1993	10.0	140	0.5239	0.8364	0.88
0.2025	11.0	154	0.5181	0.8545	0.8919
0.1417	12.0	168	0.6867	0.8364	0.8861
0.1162	13.0	182	0.7050	0.8364	0.8861
0.1124	14.0	196	0.5492	0.8545	0.8889
0.1183	15.0	210	0.7322	0.8545	0.8974
0.0693	16.0	224	0.7734	0.8545	0.8974
0.0625	17.0	238	0.6903	0.8545	0.8947
0.0803	18.0	252	0.7115	0.8727	0.9091
0.0354	19.0	266	0.7116	0.8364	0.88
0.0486	20.0	280	0.7316	0.8545	0.8919

Framework versions

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