bangla-bert-base-MLTC-BB2

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

• Loss: 0.3996
• F1: 0.8514
• F1 Weighted: 0.8510
• Roc Auc: 0.8483
• Accuracy: 0.5373
• Hamming Loss: 0.1517
• Jaccard Score: 0.7412
• Zero One Loss: 0.4627

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	F1	F1 Weighted	Roc Auc	Accuracy	Hamming Loss	Jaccard Score	Zero One Loss
0.3931	1.0	73	0.3859	0.8319	0.8312	0.8310	0.5244	0.1690	0.7123	0.4756
0.3635	2.0	146	0.3638	0.8540	0.8546	0.8528	0.5707	0.1472	0.7453	0.4293
0.2688	3.0	219	0.3676	0.8483	0.8465	0.8425	0.5553	0.1575	0.7366	0.4447
0.1841	4.0	292	0.3600	0.8562	0.8558	0.8534	0.5758	0.1465	0.7486	0.4242
0.1701	5.0	365	0.3677	0.8568	0.8565	0.8528	0.5681	0.1472	0.7495	0.4319
0.1499	6.0	438	0.3701	0.8551	0.8541	0.8534	0.5630	0.1465	0.7469	0.4370
0.1034	7.0	511	0.3825	0.8582	0.8576	0.8540	0.5527	0.1459	0.7516	0.4473
0.0793	8.0	584	0.4103	0.8443	0.8428	0.8393	0.5167	0.1607	0.7306	0.4833
0.0801	9.0	657	0.3960	0.8484	0.8484	0.8470	0.5373	0.1530	0.7367	0.4627
0.0812	10.0	730	0.3996	0.8514	0.8510	0.8483	0.5373	0.1517	0.7412	0.4627

Framework versions

• Transformers 4.41.2
• Pytorch 2.1.2
• Datasets 2.19.2
• Tokenizers 0.19.1