Pegasus-x-base-govreport-12288-1024-numepoch-5

This model is a fine-tuned version of google/pegasus-x-base on the govreport-summarization dataset. It achieves the following results on the evaluation set:

• Loss: 1.6740

Evaluation Score

For test dataset

'ROUGE':

{ 'rouge1': 0.4861,
'rouge2': 0.2067,
'rougeL': 0.2446,
'rougeLsum': 0.2444
}

'BERT_SCORE'
{'f1': 0.8551,
'precision': 0.8583,
'recall': 0.852
}

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: 0.0001
• train_batch_size: 1
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 64
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss
3.0173	0.07	20	2.6677
2.5674	0.15	40	2.2993
2.3013	0.22	60	2.1024
2.2145	0.29	80	1.9833
2.1191	0.37	100	1.9383
2.0709	0.44	120	1.8815
2.0287	0.51	140	1.8623
2.003	0.58	160	1.8467
1.9842	0.66	180	1.8314
1.9603	0.73	200	1.8307
1.9493	0.8	220	1.8157
1.9631	0.88	240	1.7919
1.9332	0.95	260	1.7919
1.9123	1.02	280	1.7836
1.887	1.1	300	1.7672
1.8743	1.17	320	1.7629
1.8412	1.24	340	1.7566
1.8508	1.32	360	1.7410
1.8564	1.39	380	1.7403
1.8686	1.46	400	1.7393
1.8881	1.53	420	1.7420
1.8629	1.61	440	1.7367
1.8683	1.68	460	1.7288
1.833	1.75	480	1.7300
1.8621	1.83	500	1.7208
1.8622	1.9	520	1.7211
1.8147	1.97	540	1.7158
1.8161	2.05	560	1.7117
1.8239	2.12	580	1.7090
1.8185	2.19	600	1.7100
1.8605	2.27	620	1.7057
1.7919	2.34	640	1.6996
1.8026	2.41	660	1.7012
1.7785	2.48	680	1.6980
1.8296	2.56	700	1.6941
1.802	2.63	720	1.6944
1.7783	2.7	740	1.6927
1.7998	2.78	760	1.6922
1.8128	2.85	780	1.6890
1.7762	2.92	800	1.6909
1.7631	3.0	820	1.6959
1.8191	3.07	840	1.6823
1.795	3.14	860	1.6873
1.7587	3.22	880	1.6850
1.8091	3.29	900	1.6828
1.7617	3.36	920	1.6860
1.7933	3.43	940	1.6796
1.8041	3.51	960	1.6805
1.7596	3.58	980	1.6855
1.7518	3.65	1000	1.6791
1.7384	3.73	1020	1.6795
1.7855	3.8	1040	1.6784
1.7938	3.87	1060	1.6780
1.7637	3.95	1080	1.6809
1.7914	4.02	1100	1.6779
1.7903	4.09	1120	1.6753
1.7874	4.17	1140	1.6745
1.7982	4.24	1160	1.6728
1.7709	4.31	1180	1.6761
1.7583	4.38	1200	1.6754
1.778	4.46	1220	1.6739
1.7526	4.53	1240	1.6746
1.7713	4.6	1260	1.6723
1.734	4.68	1280	1.6742
1.7498	4.75	1300	1.6737
1.751	4.82	1320	1.6730
1.7562	4.9	1340	1.6739
1.7549	4.97	1360	1.6740

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu117
• Datasets 2.13.1
• Tokenizers 0.13.3