results

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

• Loss: 2.6147
• Bleu: 0.2098

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: 0.0002
• train_batch_size: 16
• eval_batch_size: 16
• 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: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
No log	1.0	414	1.5123	0.1913
1.8064	2.0	828	1.5254	0.2027
1.1389	3.0	1242	1.5991	0.1996
0.771	4.0	1656	1.7113	0.2013
0.5395	5.0	2070	1.8532	0.2015
0.5395	6.0	2484	1.9543	0.1955
0.3855	7.0	2898	2.0825	0.2021
0.2794	8.0	3312	2.1461	0.1981
0.2203	9.0	3726	2.2048	0.2049
0.1833	10.0	4140	2.2585	0.1997
0.154	11.0	4554	2.2970	0.2018
0.154	12.0	4968	2.3328	0.2013
0.1266	13.0	5382	2.3380	0.2012
0.1054	14.0	5796	2.4021	0.2026
0.0905	15.0	6210	2.4106	0.2005
0.0776	16.0	6624	2.4528	0.1988
0.0665	17.0	7038	2.4778	0.2017
0.0665	18.0	7452	2.5210	0.2031
0.0563	19.0	7866	2.5157	0.2029
0.0487	20.0	8280	2.5245	0.1998
0.0411	21.0	8694	2.5513	0.2016
0.0346	22.0	9108	2.5436	0.1994
0.0304	23.0	9522	2.5845	0.1996
0.0304	24.0	9936	2.5827	0.2005
0.0237	25.0	10350	2.5879	0.2067
0.0201	26.0	10764	2.5744	0.2050
0.0177	27.0	11178	2.6031	0.2091
0.0151	28.0	11592	2.5932	0.2099
0.0133	29.0	12006	2.6221	0.2105
0.0133	30.0	12420	2.6147	0.2098

Framework versions

• Transformers 4.52.2
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1