wav2vec2-tamil-binary

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

• Loss: 0.5318
• Accuracy: 0.8235
• F1: 0.7692

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
0.6791	1.0	13	0.6893	0.5686	0.0
0.6733	2.0	26	0.6779	0.5686	0.0
0.6518	3.0	39	0.6433	0.6471	0.3077
0.6022	4.0	52	0.5652	0.8235	0.7429
0.5153	5.0	65	0.5481	0.7647	0.625
0.4387	6.0	78	0.4189	0.8431	0.8000
0.3883	7.0	91	0.3900	0.8431	0.8000
0.3375	8.0	104	0.4485	0.8039	0.7368
0.2891	9.0	117	0.5500	0.7647	0.6471
0.2961	10.0	130	0.3474	0.8431	0.8000
0.231	11.0	143	0.5168	0.8235	0.7568
0.2029	12.0	156	0.3744	0.8431	0.8000
0.2102	13.0	169	0.4384	0.8235	0.7692
0.1564	14.0	182	0.4245	0.8431	0.8000
0.1565	15.0	195	0.4967	0.8235	0.7692
0.1308	16.0	208	0.5096	0.8235	0.7692
0.1578	17.0	221	0.4682	0.8235	0.7692
0.1581	18.0	234	0.5343	0.8431	0.7895
0.143	19.0	247	0.5352	0.8235	0.7692
0.1637	20.0	260	0.5318	0.8235	0.7692

Framework versions

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