iter2-undiacritized

This model is a fine-tuned version of roberta-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.1465
• Accuracy: 0.6701
• F1 Macro: 0.6695
• Composite Score: 0.6695

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP
• label_smoothing_factor: 0.2

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro	Composite Score
0.8536	1.0	2208	1.2374	0.3824	0.3265	0.2820
0.7304	2.0	4416	1.1405	0.5716	0.5581	0.5527
0.6069	3.0	6624	1.1397	0.5973	0.5873	0.5870
0.5614	4.0	8832	1.1043	0.6437	0.6441	0.6441
0.535	5.0	11040	1.1086	0.6659	0.6663	0.6663
0.5253	6.0	13248	1.1278	0.6712	0.6726	0.6726
0.5158	7.0	15456	1.1527	0.6576	0.6591	0.6591
0.5126	8.0	17664	1.1264	0.6772	0.6788	0.6788
0.505	9.0	19872	1.1422	0.6674	0.6682	0.6682
0.4998	10.0	22080	1.1301	0.6742	0.6736	0.6736
0.4962	11.0	24288	1.1465	0.6701	0.6695	0.6695

Framework versions

• Transformers 4.53.3
• Pytorch 2.6.0+cu124
• Datasets 4.4.1
• Tokenizers 0.21.2