bart-abs-1509-0313-lr-3e-05-bs-4-maxep-10

This model is a fine-tuned version of sshleifer/distilbart-xsum-12-6 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.5339
• Rouge/rouge1: 0.4605
• Rouge/rouge2: 0.2065
• Rouge/rougel: 0.3887
• Rouge/rougelsum: 0.3902
• Bertscore/bertscore-precision: 0.8931
• Bertscore/bertscore-recall: 0.8914
• Bertscore/bertscore-f1: 0.8921
• Meteor: 0.4157
• Gen Len: 37.8364

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: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge/rouge1	Rouge/rouge2	Rouge/rougel	Rouge/rougelsum	Bertscore/bertscore-precision	Bertscore/bertscore-recall	Bertscore/bertscore-f1	Meteor	Gen Len
1.2307	1.0	217	2.1312	0.4591	0.2098	0.3873	0.3883	0.8977	0.892	0.8947	0.4036	37.4909
0.888	2.0	434	2.2053	0.4516	0.2075	0.3834	0.3857	0.8957	0.8906	0.893	0.3977	35.2273
0.784	3.0	651	2.3695	0.4573	0.2136	0.3893	0.3911	0.8972	0.8904	0.8937	0.4003	34.8
0.5673	4.0	868	2.6218	0.4714	0.2118	0.3982	0.3996	0.8947	0.8924	0.8934	0.4235	39.2727
0.4163	5.0	1085	2.9151	0.4683	0.2131	0.4005	0.4023	0.8958	0.8916	0.8935	0.4129	36.7545
0.3021	6.0	1302	3.0962	0.4648	0.2045	0.3918	0.3935	0.8967	0.893	0.8947	0.4119	37.0636
0.2266	7.0	1519	3.2782	0.4639	0.2074	0.3907	0.3925	0.8942	0.8941	0.894	0.4203	38.9273
0.1684	8.0	1736	3.4198	0.4565	0.1964	0.3822	0.3841	0.8934	0.8905	0.8918	0.4035	37.1818
0.1347	9.0	1953	3.4878	0.4723	0.2189	0.3987	0.4005	0.8954	0.8957	0.8954	0.4308	39.5273
0.1124	10.0	2170	3.5339	0.4605	0.2065	0.3887	0.3902	0.8931	0.8914	0.8921	0.4157	37.8364

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1