sum_model_lr1e_3_20epoch

This model got the best result so far.

This model is a fine-tuned version of weny22/sum_model_t5_saved on the INF582-2023-24 dataset. It achieves the following results on the evaluation set:

• Loss: 1.8879
• Rouge1: 0.2188
• Rouge2: 0.0915
• Rougel: 0.181
• Rougelsum: 0.1808
• Gen Len: 18.98

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: 0.001
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
No log	1.0	335	2.1280	0.196	0.0662	0.156	0.1559	18.988
2.8114	2.0	670	2.0104	0.2004	0.0724	0.1609	0.1609	18.956
2.2319	3.0	1005	1.9785	0.2082	0.0776	0.1681	0.1681	18.964
2.2319	4.0	1340	1.9377	0.2084	0.0831	0.1703	0.1704	18.9787
2.0444	5.0	1675	1.8873	0.2107	0.0836	0.1719	0.1722	18.9813
1.9359	6.0	2010	1.8945	0.2132	0.0848	0.1736	0.1735	18.9733
1.9359	7.0	2345	1.8949	0.2135	0.0843	0.1725	0.1727	18.9627
1.8292	8.0	2680	1.8741	0.2155	0.0869	0.1762	0.1765	18.9487
1.7623	9.0	3015	1.8679	0.2154	0.0873	0.176	0.1759	18.9767
1.7623	10.0	3350	1.8627	0.2171	0.0883	0.1774	0.1775	18.9833
1.6812	11.0	3685	1.8617	0.217	0.0877	0.176	0.1759	18.9827
1.6331	12.0	4020	1.8572	0.2154	0.088	0.1756	0.1757	18.982
1.6331	13.0	4355	1.8645	0.2175	0.0895	0.178	0.178	18.972
1.5737	14.0	4690	1.8707	0.2168	0.0877	0.1761	0.1761	18.978
1.5326	15.0	5025	1.8764	0.2204	0.09	0.1805	0.1804	18.9827
1.5326	16.0	5360	1.8746	0.2196	0.0916	0.1804	0.1804	18.9767
1.4881	17.0	5695	1.8734	0.2195	0.0924	0.1804	0.1806	18.9867
1.4631	18.0	6030	1.8869	0.219	0.091	0.1802	0.1802	18.972
1.4631	19.0	6365	1.8886	0.2201	0.092	0.1819	0.1819	18.9847
1.4345	20.0	6700	1.8879	0.2188	0.0915	0.181	0.1808	18.98

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2