manual_chakma_sagorBangla

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: 2.8966

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
4.9918	1.0	124	3.8939
4.1444	2.0	248	3.6157
3.8256	3.0	372	3.4138
3.6106	4.0	496	3.2417
3.497	5.0	620	3.2316
3.3379	6.0	744	3.3402
3.2328	7.0	868	3.1591
3.1481	8.0	992	3.0594
3.0173	9.0	1116	3.1641
2.9797	10.0	1240	3.1307
2.891	11.0	1364	2.9460
2.8874	12.0	1488	3.1390
2.7933	13.0	1612	2.9611
2.7871	14.0	1736	3.0125
2.7203	15.0	1860	2.9763
2.6993	16.0	1984	2.9991
2.6842	17.0	2108	3.0116
2.6422	18.0	2232	3.0069
2.6128	19.0	2356	2.8798
2.6222	20.0	2480	2.8966

Framework versions

• Transformers 4.56.1
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.0