gemini_chakma_roberta

This model is a fine-tuned version of FacebookAI/xlm-roberta-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.2323

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
3.1922	1.0	205	3.3312
2.8762	2.0	410	3.1542
2.6543	3.0	615	3.0181
2.5254	4.0	820	2.8706
2.4038	5.0	1025	2.7951
2.325	6.0	1230	2.7440
2.2749	7.0	1435	2.6188
2.1967	8.0	1640	2.6125
2.1665	9.0	1845	2.4628
2.1042	10.0	2050	2.4725
2.0445	11.0	2255	2.3773
2.0271	12.0	2460	2.3718
2.0039	13.0	2665	2.3011
1.9585	14.0	2870	2.3232
1.961	15.0	3075	2.2512
1.914	16.0	3280	2.2426
1.9351	17.0	3485	2.2657
1.8987	18.0	3690	2.2187
1.8596	19.0	3895	2.2450
1.8777	20.0	4100	2.2323

Framework versions

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