indicbert-hatespeechdetection-tamil

This model is a fine-tuned version of ai4bharat/IndicBERTv2-MLM-only on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3477
• Accuracy: 0.9216
• F1: 0.9091

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• 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
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.6787	1.0	13	0.6313	0.6471	0.3571
0.5693	2.0	26	0.5115	0.7843	0.6667
0.4117	3.0	39	0.3694	0.8431	0.8095
0.2801	4.0	52	0.3536	0.8431	0.8000
0.2099	5.0	65	0.3527	0.8627	0.8293
0.141	6.0	78	0.2759	0.8824	0.8571
0.0855	7.0	91	0.2467	0.8824	0.8571
0.0528	8.0	104	0.3379	0.8627	0.8205
0.035	9.0	117	0.1741	0.9412	0.9333
0.0299	10.0	130	0.3975	0.8824	0.85
0.0123	11.0	143	0.2882	0.9216	0.9091
0.0077	12.0	156	0.3869	0.9020	0.8780
0.0034	13.0	169	0.2488	0.9216	0.9048
0.0019	14.0	182	0.2398	0.9412	0.9333
0.0018	15.0	195	0.3282	0.9216	0.9048
0.0015	16.0	208	0.3737	0.9020	0.8780
0.0015	17.0	221	0.3222	0.9216	0.9091
0.0012	18.0	234	0.3025	0.9216	0.9091
0.0016	19.0	247	0.3345	0.9216	0.9091
0.0012	20.0	260	0.3477	0.9216	0.9091

Framework versions

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