gemini_chakma_bert

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: 1.7301

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
2.5731	1.0	228	2.7399
2.2245	2.0	456	2.4805
2.0515	3.0	684	2.3670
1.9288	4.0	912	2.2668
1.8429	5.0	1140	2.1341
1.7925	6.0	1368	2.1439
1.7298	7.0	1596	2.0550
1.6792	8.0	1824	2.0018
1.625	9.0	2052	1.9309
1.5888	10.0	2280	1.9261
1.5427	11.0	2508	1.8502
1.5201	12.0	2736	1.8298
1.4936	13.0	2964	1.8438
1.4709	14.0	3192	1.7637
1.4441	15.0	3420	1.7729
1.4201	16.0	3648	1.7761
1.4208	17.0	3876	1.7413
1.4038	18.0	4104	1.7010
1.3808	19.0	4332	1.7308
1.3808	20.0	4560	1.7301

Framework versions

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