chakma-bert

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

• Loss: 1.8670

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• 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: 18
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
3.1287	1.0	59	2.8297
2.7974	2.0	118	2.5401
2.5044	3.0	177	2.4909
2.432	4.0	236	2.2127
2.3317	5.0	295	2.2474
2.1859	6.0	354	2.2014
2.1582	7.0	413	2.1784
2.0694	8.0	472	2.0472
2.0662	9.0	531	2.0296
2.0164	10.0	590	2.0319
1.9955	11.0	649	1.9045
1.8953	12.0	708	1.9912
1.9059	13.0	767	1.9668
1.8877	14.0	826	1.8459
1.8675	15.0	885	1.8975
1.8644	16.0	944	1.8578
1.8482	17.0	1003	1.9558
1.8276	18.0	1062	1.8670

Framework versions

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