BARTBana_Translation_v5

This model is a fine-tuned version of IAmSkyDra/BARTBana_v5 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5493
• Sacrebleu: 8.2109

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: 64
• eval_batch_size: 64
• 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
• num_epochs: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Sacrebleu
0.8112	1.0	742	0.7198	1.9252
0.7466	2.0	1484	0.6679	3.0362
0.6719	3.0	2226	0.6347	4.1068
0.6522	4.0	2968	0.6128	4.9362
0.6111	5.0	3710	0.5966	5.4351
0.5941	6.0	4452	0.5835	5.9868
0.5613	7.0	5194	0.5753	6.4039
0.5428	8.0	5936	0.5684	6.6599
0.5315	9.0	6678	0.5607	6.8644
0.5132	10.0	7420	0.5572	7.1633
0.4958	11.0	8162	0.5534	7.2500
0.4849	12.0	8904	0.5544	7.5064
0.4731	13.0	9646	0.5502	7.6249
0.4624	14.0	10388	0.5508	7.5880
0.4482	15.0	11130	0.5503	7.7981
0.4434	16.0	11872	0.5488	7.9009
0.43	17.0	12614	0.5488	7.8955
0.425	18.0	13356	0.5500	8.0053
0.4177	19.0	14098	0.5460	8.1097
0.4121	20.0	14840	0.5489	8.0999
0.4124	21.0	15582	0.5498	8.1294
0.4046	22.0	16324	0.5492	8.2079
0.4011	23.0	17066	0.5493	8.2109
0.4049	24.0	17808	0.5511	8.1965
0.3964	25.0	18550	0.5513	8.1733

Framework versions

• Transformers 4.48.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0