bangla-fake-news

This model is a fine-tuned version of csebuetnlp/banglabert on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4802
• Accuracy: 0.9219
• F1: 0.9533

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: 16
• eval_batch_size: 16
• 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: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.3343	0.2998	950	0.3675	0.9148	0.9488
0.3942	0.5996	1900	0.3998	0.9311	0.9593
0.3804	0.8993	2850	0.4338	0.9322	0.9601
0.3698	1.1991	3800	0.4205	0.9301	0.9587
0.3444	1.4989	4750	0.4973	0.9217	0.9534
0.3475	1.7987	5700	0.4072	0.9292	0.9582
0.3436	2.0985	6650	0.3921	0.9094	0.9450
0.3017	2.3982	7600	0.4264	0.9078	0.9441
0.3115	2.6980	8550	0.4774	0.9187	0.9512
0.31	2.9978	9500	0.4802	0.9219	0.9533

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Tokenizers 0.21.0