wav2vec2-telugu-binary

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

• Loss: 0.6840
• Accuracy: 0.6727
• F1: 0.7188

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
0.6817	1.0	14	0.6941	0.3636	0.0
0.6793	2.0	28	0.6921	0.5273	0.5185
0.6725	3.0	42	0.6872	0.6182	0.7640
0.6591	4.0	56	0.6743	0.5818	0.6567
0.6131	5.0	70	0.6616	0.6	0.6857
0.5472	6.0	84	0.6733	0.5818	0.6462
0.4911	7.0	98	0.6922	0.6	0.6667
0.467	8.0	112	0.7286	0.5818	0.6462
0.471	9.0	126	0.7040	0.6182	0.6866
0.422	10.0	140	0.7001	0.6545	0.7246
0.3961	11.0	154	0.6830	0.6364	0.6875
0.3804	12.0	168	0.6636	0.6545	0.7077
0.3753	13.0	182	0.6657	0.7273	0.7945
0.3482	14.0	196	0.7015	0.6545	0.6984
0.317	15.0	210	0.6906	0.6545	0.7077
0.3091	16.0	224	0.6977	0.6545	0.6984
0.2797	17.0	238	0.6670	0.6727	0.7188
0.2866	18.0	252	0.6562	0.7091	0.7576
0.2913	19.0	266	0.6849	0.6727	0.7188
0.2755	20.0	280	0.6840	0.6727	0.7188

Framework versions

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