T5_base_title_v3

This model is a fine-tuned version of t5-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.8544
• Rouge1: 0.4104
• Rouge2: 0.212
• Rougel: 0.3579
• Rougelsum: 0.3576
• Gen Len: 16.585

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
2.2216	1.0	500	1.9241	0.3884	0.1992	0.3375	0.3371	16.168
1.9818	2.0	1000	1.8699	0.391	0.2002	0.342	0.341	16.386
1.8876	3.0	1500	1.8377	0.3972	0.2033	0.3447	0.3434	16.635
1.805	4.0	2000	1.8202	0.3981	0.2061	0.3482	0.3477	16.213
1.7422	5.0	2500	1.8180	0.395	0.2051	0.345	0.3445	16.74
1.6919	6.0	3000	1.8154	0.4042	0.2091	0.3526	0.3519	16.197
1.6426	7.0	3500	1.8160	0.4048	0.2094	0.3509	0.3506	16.546
1.5982	8.0	4000	1.8224	0.4088	0.2131	0.3556	0.3552	16.516
1.5615	9.0	4500	1.8219	0.4079	0.2121	0.3549	0.3545	16.474
1.5304	10.0	5000	1.8235	0.4059	0.2128	0.3548	0.3547	16.498
1.4996	11.0	5500	1.8299	0.4098	0.211	0.3569	0.3564	16.378
1.4735	12.0	6000	1.8349	0.4108	0.2129	0.3576	0.3572	16.672
1.4502	13.0	6500	1.8373	0.4103	0.2132	0.3566	0.3565	16.68
1.426	14.0	7000	1.8434	0.4092	0.2104	0.3559	0.3554	16.669
1.4137	15.0	7500	1.8442	0.4117	0.212	0.358	0.3576	16.547
1.401	16.0	8000	1.8503	0.4109	0.2126	0.3569	0.3567	16.552
1.3901	17.0	8500	1.8517	0.4115	0.2146	0.3601	0.36	16.553
1.3817	18.0	9000	1.8539	0.4104	0.2118	0.3572	0.3573	16.588
1.3696	19.0	9500	1.8553	0.4103	0.2126	0.3581	0.3578	16.568
1.369	20.0	10000	1.8544	0.4104	0.212	0.3579	0.3576	16.585

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.2
• Datasets 2.1.0
• Tokenizers 0.15.1