mbert-hatespeechdetection-telugu

This model is a fine-tuned version of google-bert/bert-base-multilingual-cased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4645
• Accuracy: 0.9286
• F1: 0.9459

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
0.6103	1.0	14	0.5045	0.6429	0.7826
0.5228	2.0	28	0.4161	0.7679	0.8354
0.4483	3.0	42	0.3746	0.8393	0.8571
0.4312	4.0	56	0.3793	0.7857	0.8
0.3572	5.0	70	0.3443	0.8571	0.8824
0.2675	6.0	84	0.2686	0.875	0.8923
0.2355	7.0	98	0.3259	0.8393	0.8571
0.1627	8.0	112	0.2204	0.875	0.8986
0.1151	9.0	126	0.2648	0.8929	0.9091
0.1126	10.0	140	0.3640	0.8393	0.8657
0.0257	11.0	154	0.2720	0.9107	0.9333
0.0065	12.0	168	0.4303	0.8929	0.9211
0.0303	13.0	182	0.4108	0.9286	0.9459
0.0016	14.0	196	0.4924	0.9107	0.9333
0.0126	15.0	210	0.4652	0.9107	0.9315
0.0047	16.0	224	0.4827	0.9107	0.9315
0.0009	17.0	238	0.4614	0.9286	0.9459
0.0008	18.0	252	0.4613	0.9286	0.9459
0.0007	19.0	266	0.4634	0.9286	0.9459
0.0007	20.0	280	0.4645	0.9286	0.9459

Framework versions

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